Direct access to the Project Report
In this project we combined two neural networks to perform object recognition and depths estimation tasks simultaneously.
Official repository: (https://github.com/Nadiam75/joint_object_detection_depth_estimation)
Pretrained Depth Estimation: (https://github.innominds.com/karoly-hars/DE_resnet_unet_hyb)
Pretrained Yolo_V5: (https://github.com/ultralytics/yolov5)
Pretrained Yolo_V2: (https://pjreddie.com/darknet/yolov2)
Dataset Used to Train the Network (https://cs.nyu.edu/~silberman/datasets/nyu_depth_v2.html
- Python
- Pytorch
- Opencv-Python
- Matplotlib
- h5py
- PIL
- scipy
- tensorflow
- torchvision
Contains detailes on the implementation of the three structures implemented, our telegram bot and the webApp.
DL_Project.ipynb
joint object detection and depth estimation using pretrained YOLO_V5 and Pretrained Depth Estimation
DL_YOLOv2_ResnetUnetHybrid.ipynb
joint object detection and depth estimation using pretrained YOLO_V2 and Pretrained Depth Estimation
trained_depth_yolo_v5
joint object detection and depth estimation using pretrained YOLO_V2 and trainin Depth Estimation on NYU dataset
Telgeram bot available at: @DL_Sharif_Project_bot
Our telegram bot is capabale of detecting objects and estimating their corresponding depths closer or farther from 2, 3 or 4 meters, this threshold can be changed according to the users needs!
In order to install the dependencies run the following commands in shell
pip install -r requirements.txtTo Start the Server Run the following command:
python manage.py runserver
Here are some of our training results on TEST DATASET (depth estimation model has been trained for 30 epochs on NYU):
A short video containing details on how to use the web application has also been uploaded.
Email: [email protected], [email protected]
Welcome for any discussions!