Updated the code that allows to save the cropped image in result_img directory #7282

shubham-shahh · 2021-01-24T09:53:59Z

Bug Fix

The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' and the code to save the image in result_img dir isn't
updated based on the above conversion which was causing errors while recompiling the code.
So, I've updated that, and now one can store the detected images in result_img dir. I've also updated the Readme.md based
on that
Fixes: How to get co-ordinates and save image of the objects #7276

Feature Addition

I've added a new feature with which a user can run inference on all the images in the mentioned folder. I've also updated the readme with the information on how to use this feature.

README.md

cenit · 2021-01-24T14:41:40Z

I have updated your initial message and reopened initial issue. The magic token "Fixes: #issue" will close the issue if/when this PR is merged automatically

shubham-shahh · 2021-01-24T14:54:02Z

I have updated your initial message and reopened initial issue. The magic token "Fixes: #issue" will close the issue if/when this PR is merged automatically

okay, thanks :)

README.md

committed the changes suggested Co-authored-by: Stefano Sinigardi <[email protected]>

cenit · 2021-01-25T14:50:10Z

@AlexeyAB what's your opinion on this PR? It fixes lines of code that are commented, as such they do not break nor fix anything. But, if the user was to uncomment those lines, they were broken and now ok

…erminating)

shubham-shahh · 2021-02-07T13:52:36Z

The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' and the code to save the image in result_img dir isn't
updated based on the above conversion which was causing errors while recompiling the code.
So, I've updated that, and now one can store the detected images in result_img dir. I've also updated the Readme.md based
on that
Fixes: How to get co-ordinates and save image of the objects #7276

I've added a new feature with which a user can run inference on all the images in the mentioned folder. I've also updated the readme with the information on how to use this feature.

@cenit sir please review my recent changes, in which I've added this feature.
Thanks.

shubham-shahh · 2021-02-07T14:07:52Z

README.md

@@ -288,6 +288,8 @@ On Linux find executable file `./darknet` in the root directory, while on Window
 * To calculate anchors: `darknet.exe detector calc_anchors data/obj.data -num_of_clusters 9 -width 416 -height 416`
 * To check accuracy mAP@IoU=50: `darknet.exe detector map data/obj.data yolo-obj.cfg backup\yolo-obj_7000.weights`
 * To check accuracy mAP@IoU=75: `darknet.exe detector map data/obj.data yolo-obj.cfg backup\yolo-obj_7000.weights -iou_thresh 0.75`
+* To save the cropped image of the object detected: [`uncomment the following lines of code`](https://github.com/AlexeyAB/darknet/blob/master/src/image_opencv.cpp#L980-L996)
+* To run inference on all the images in a folder: `./darknet detector test ./cfg/coco.data ./cfg/yolov4.cfg ./yolov4.weights -i 0 -thresh 0.25 --folder_inference ./result_img/` here `./result_img/` is the path of the folder in which the inference has to be done. the folder name can be anything, as long as it exists. After the inference the images will be moved to `./processed_files/`. To change the folder name or path, the changes cane be done [`here`](https://github.com/shubham-shahh/darknet/blob/ebc7b6f3d575402fbcd67cf214243d437f580fe1/src/detector.c#L1895).


updated the readme.md

As always, please use links to alexeyab's repo, without any commit id reference but pointing to the latest master trunk

okay, sir, I'll do that

Sir, I've updated the link but it is broken until the PR is merged @cenit

Yes of course link is broken until merged, but this is a requirement for being considered for merging!

include/darknet.h

shubham-shahh · 2021-02-07T14:10:30Z

src/detector.c

-            if (lk.type == YOLO || lk.type == GAUSSIAN_YOLO || lk.type == REGION) {
-                l = lk;
-                printf(" Detection layer: %d - type = %d \n", k, l.type);
+    if(folder_inference==NULL){


made changes in such a way that the original functionality is not changed by any means

AlexeyAB · 2021-02-07T15:16:47Z

But why is
./darknet detector test ./cfg/coco.data ./cfg/yolov4.cfg ./yolov4.weights -i 0 -thresh 0.25 --folder_inference ./result_img/
better than
darknet.exe detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < data/train.txt > result.txt ?

This < data/train.txt is more flexible, since you can put either all images from folder to the data/train.txt or only some images.

shubham-shahh · 2021-02-07T17:24:46Z

But why is
./darknet detector test ./cfg/coco.data ./cfg/yolov4.cfg ./yolov4.weights -i 0 -thresh 0.25 --folder_inference ./result_img/
better than
darknet.exe detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < data/train.txt > result.txt ?

This < data/train.txt is more flexible, since you can put either all images from folder to the data/train.txt or only some images.

Hello sir, I hope you are doing well.
darknet.exe detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < data/train.txt > result.txt this command works well when the user knows the list of images in a folder and if no new images are added or subtracted from the folder because the program ends when the inference on all the images is done

./darknet detector test ./cfg/coco.data ./cfg/yolov4.cfg ./yolov4.weights -i 0 -thresh 0.25 --folder_inference ./result_img/ this command works well when new images are added constantly to a folder, as this program is non terminating and keeps looking for images in the folder. A common use case could be if new images are added to a folder from some other process and the user wants to do further inference on all the images in that folder. in this case new images are constantly added to the folder hence a user cannot make a list of it.
one more example I can think of is if a user made an API to use their custom trained YOLO model, in that case, the images uploaded by the client can be stored in a folder on which the inference is done and the results are sent back

thanks for this awesome project @AlexeyAB

AlexeyAB · 2021-02-07T20:20:06Z

I think it is not worth to complicate the code due to the functionality that is easily achieved by these functions.

Linux:

ls /home/user/*.jpg > filelist.txt

./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt

Windows:

dir C:\images /b /s > "filelist.txt"

darknet.exe detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt

shubham-shahh · 2021-02-07T23:30:34Z

I think it is not worth to complicate the code due to the functionality that is easily achieved by these functions.

Linux:

ls /home/user/*.jpg > filelist.txt

./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt

Windows:

dir C:\images /b /s > "filelist.txt"

darknet.exe detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt

Sir, The procedure you mentioned is still terminating and the only way I can think of to make it non terminating is by creating a shell script that creates .txt files after every regular interval or to edit the code. In this approach, the user also has to make another .txt file to keep a track of images on which the inference has been done or delete the images or move them which again requires editing the code or putting that in a shell script. with this approach,
./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt has to be executed again and again which means the user has to load the model in GPU memory at regular intervals which is time-consuming and resource intensive.
Regarding the complexity of code, I've added a simple if statement if the user wants to use this functionality this block of code will execute. and if the user doesn't use the functionality the code would execute the way it used to be.

rickkcir · 2021-10-17T15:27:03Z

./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt

NB!

shubham-shahh · 2021-10-17T17:33:01Z

./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt

NB!

What are you trying to say? Can you please clearly describe your issue?

rickkcir · 2021-10-18T08:21:50Z

./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt
NB!

What are you trying to say? Can you please clearly describe your issue?

I want to get the coordinates,then I use this commad, but I found the file only saved relative coordinates,while I want to save the absolute coordinates, do you know how to realize it? thanks

rg321 · 2021-11-02T17:19:02Z

Hi, can you try following code
in file src/image_opencv.cpp, show_img is Mat pointer.
Just crop desired portion out of Mat.

// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);

pull request created for the same
#8197

shubham-shahh · 2021-11-02T17:28:48Z

./darknet detector test cfg/coco.data cfg/yolov4.cfg yolov4.weights -ext_output -dont_show -out result.json < filelist.txt > result.txt
NB!

What are you trying to say? Can you please clearly describe your issue?

I want to get the coordinates,then I use this commad, but I found the file only saved relative coordinates,while I want to save the absolute coordinates, do you know how to realize it? thanks

You can get the absolute co ordinates since you already know the original image size

shubham-shahh · 2021-11-02T17:29:53Z

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.

// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);

pull request created for the same #8197

Can you clearly mention what exactly you're expecting?

Thanks

shubham-shahh

@cenit what does this commit do?

rg321 · 2021-11-16T19:41:55Z

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?

Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -:
cv::Mat copy_image = (*show_img)(rect)
rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -:
cv::imwrite(image_name, copy_image)
Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

shubham-shahh · 2021-11-17T02:32:48Z

src/image_opencv.cpp

-                //    if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels);
-                //    cvCopy(show_img, copy_img, 0);
+                //    if(copy_img == NULL) copy_img = cvCreateImage(cvSize(ipltypeimg->width, ipltypeimg->height), ipltypeimg->depth, ipltypeimg->nChannels);
+                //    cvCopy(ipltypeimg, copy_img, 0);


It is solved here @rg321

shubham-shahh · 2021-11-17T02:34:20Z

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

H

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

Hi, this PR also implements some more features along with solving that bug. That bug is simply solved by converting the image to ipltype, I tagged you where I have done that conversion. Rest changes of this PR are made to implement several other features

Thanks

rg321 · 2021-11-24T08:35:16Z

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

H

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

Hi, this PR also implements some more features along with solving that bug. That bug is simply solved by converting the image to ipltype, I tagged you where I have done that conversion. Rest changes of this PR are made to implement several other features

Thanks

In that case, you may either exactly mention in title of this pull request or you may want to create different pull request for the bug and a different one for features implemented. That would provide more clarity to your pull request. If possible, provide appropriate title to your pull request, what bug it solves or what feature it implements and add changes necessary for that only in that particular pull request. Else, as I mentioned above, for the problem mentioned in the title, my changes are good enough.

shubham-shahh · 2021-11-24T08:50:23Z

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

H

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

i

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

H

Hi, can you try following code in file src/image_opencv.cpp, show_img is Mat pointer. Just crop desired portion out of Mat.
// you should create directory: result_img static int copied_frame_id = -1; // static IplImage* copy_img = NULL; if (copied_frame_id != frame_id) { copied_frame_id = frame_id; // if(copy_img == NULL) copy_img = cvCreateImage(cvSize(show_img->width, show_img->height), show_img->depth, show_img->nChannels); // cvCopy(show_img, copy_img, 0); } static int img_id = 0; img_id++; char image_name[1024]; sprintf(image_name, "result_img/img_%d_%d_%d_%s.jpg", frame_id, img_id, class_id, names[class_id]); CvRect rect = cvRect(pt1.x, pt1.y, pt2.x - pt1.x, pt2.y - pt1.y); cv::Mat copy_image = (*show_img)(rect); cv::imwrite(image_name, copy_image); // cvSetImageROI(copy_img, rect); // cvSaveImage(image_name, copy_img, 0); // cvResetImageROI(copy_img);
pull request created for the same #8197

Can you clearly mention what exactly you're expecting?
Thanks

@shubham-shahh as you mentioned at top of this page, you are trying to crop detections out of image. I just provided a simple way of doing that. As you pointed out, " The type of 'show_img' has been changed from 'iplimage' to 'cv::Mat' ", so just crop the detected area out of original image. For that I have just added following single line -: cv::Mat copy_image = (*show_img)(rect) rect is the rectange of detected object. We just crop that area from show_img and name it copy_image. Now save copy_image and it is cropped detectetion out of original image. To save copy_image, I use following code -: cv::imwrite(image_name, copy_image) Thats it, problem solved. You don't have to worry that " 'show_img' has been changed from 'iplimage' to 'cv::Mat' ".

Hi, this PR also implements some more features along with solving that bug. That bug is simply solved by converting the image to ipltype, I tagged you where I have done that conversion. Rest changes of this PR are made to implement several other features
Thanks

In that case, you may either exactly mention in title of this pull request or you may want to create different pull request for the bug and a different one for features implemented. That would provide more clarity to your pull request. If possible, provide appropriate title to your pull request, what bug it solves or what feature it implements and add changes necessary for that only in that particular pull request. Else, as I mentioned above, for the problem mentioned in the title, my changes are good enough.

Hi,
you can check the description, I have clearly mentioned what bug it solves and which feature it implements

thanks

cenit · 2021-11-30T08:39:24Z

src/detector.c

@@ -8,6 +8,7 @@
 #include "box.h"
 #include "demo.h"
 #include "option_list.h"
+#include <dirent.h>


PR is using dirent.h and code that is not cross-compatible, so is unacceptable as is

I can add alternatives to dirent @cenit

can you please tag what part of the code which is causing issue? I will work on it. plus this is a really old PR it needs rebasing as well

I rebased it on latest master for you before commenting the issue

just rewrite the code in this file avoiding dirent.h inclusion (not available on windows)

okay, thanks for pointing out the issues, I'll work on them

cenit · 2021-11-30T09:04:06Z

src/detector.c

+        DIR *folder;
+        struct dirent *entry;


these two lines, for example, depend on dirent.h being included, which is incompatible with our aim of being 100% windows compatible

okay, I noted it, I'll apply required changes

what's your opinion for Boost filesystem? I don't have much experience in building projects in windows. all your help is appreciated

if you can use std::filesystem it would be much better, so that we can avoid depending on another external lib

Okay, I'll look into it
thanks

Hi, @cenit I was planning to fix this PR, I just wanted to know if there are any other changes to make other than filesystem?

shubham-shahh · 2023-10-02T02:29:13Z

Hi @cenit looks like you've fixed the issue related to building in windows, can we merge this PR into master?

shubham-shahh mentioned this pull request Jan 24, 2021

How to get co-ordinates and save image of the objects #7276

Open

cenit reviewed Jan 24, 2021

View reviewed changes

README.md Outdated Show resolved Hide resolved

cenit reviewed Jan 24, 2021

View reviewed changes

README.md Outdated Show resolved Hide resolved

shubham-shahh and others added 6 commits January 24, 2021 20:26

rectified the bug in src/image_opencv.cpp

2d3bc5e

added relevant comment

95ec9bd

updated the Readme.md based on the changes applied

a465e8a

fixed a typo

e5a6346

fixed the link in Readme.md

29eb986

Update README.md

43e1833

committed the changes suggested Co-authored-by: Stefano Sinigardi <[email protected]>

shubham-shahh added 4 commits February 7, 2021 18:39

added a feature to run inference on all the images in a folder (non-t…

317a898

…erminating)

updated the readme.md and removed unwanted code

ebc7b6f

updated the readme.md

661010d

removed unnecessary code

e1b46be

shubham-shahh commented Feb 7, 2021

View reviewed changes

include/darknet.h Show resolved Hide resolved

shubham-shahh commented Feb 7, 2021

View reviewed changes

updated readme.md

cb66764

shubham-shahh mentioned this pull request Feb 9, 2021

Run YoloV3 detections on thousands of images and save outputs? pjreddie/darknet#723

Closed

Merge branch 'master' into master

74f8fbe

rg321 mentioned this pull request Nov 2, 2021

changes to crop detections in video #8197

Open

shubham-shahh commented Nov 2, 2021

View reviewed changes

shubham-shahh commented Nov 17, 2021

View reviewed changes

Merge branch 'master' into master

7d1a1d6

cenit reviewed Nov 30, 2021

View reviewed changes

cenit added 4 commits September 1, 2023 01:48

Merge branch 'master' into pr/7282

1ff1260

restore building compatibility

6b703cb

Merge branch 'master' into pr/7282

37b990a

Restore build

4aad894

Updated the code that allows to save the cropped image in result_img directory #7282

Are you sure you want to change the base?

Updated the code that allows to save the cropped image in result_img directory #7282

Conversation

shubham-shahh commented Jan 24, 2021 • edited Loading

Bug Fix

Feature Addition

cenit commented Jan 24, 2021

shubham-shahh commented Jan 24, 2021

cenit commented Jan 25, 2021

shubham-shahh commented Feb 7, 2021

Choose a reason for hiding this comment

cenit Feb 7, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

AlexeyAB commented Feb 7, 2021

shubham-shahh commented Feb 7, 2021 • edited Loading

AlexeyAB commented Feb 7, 2021

shubham-shahh commented Feb 7, 2021 • edited Loading

rickkcir commented Oct 17, 2021

shubham-shahh commented Oct 17, 2021 • edited Loading

rickkcir commented Oct 18, 2021

rg321 commented Nov 2, 2021

shubham-shahh commented Nov 2, 2021

shubham-shahh commented Nov 2, 2021

shubham-shahh left a comment

Choose a reason for hiding this comment

rg321 commented Nov 16, 2021 • edited Loading

Choose a reason for hiding this comment

shubham-shahh commented Nov 17, 2021

rg321 commented Nov 24, 2021

shubham-shahh commented Nov 24, 2021

Choose a reason for hiding this comment

shubham-shahh Nov 30, 2021 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

shubham-shahh commented Oct 2, 2023

shubham-shahh commented Jan 24, 2021 •

edited

Loading

cenit Feb 7, 2021 •

edited

Loading

shubham-shahh commented Feb 7, 2021 •

edited

Loading

shubham-shahh commented Feb 7, 2021 •

edited

Loading

shubham-shahh commented Oct 17, 2021 •

edited

Loading

rg321 commented Nov 16, 2021 •

edited

Loading

shubham-shahh Nov 30, 2021 •

edited

Loading