Welcome to my first Machine learning Computer Vision project. In this project I'd like to practice a problem of object detection. I will train a Neural Network of an architecture called Faster R-CNN. Plan is to be able to detect and draw bounding boxes around tomatoes in images. Here we go.
Progress: I followed the guide from this article https://towardsdatascience.com/faster-r-cnn-object-detection-implemented-by-keras-for-custom-data-from-googles-open-images-125f62b9141a Which holds all training data needed. And besides those, A few more files are needed to be able to run the sequence, these are the vgg16 models that are pre-trained. These are available to download through a simple google search. The files are:
- model_frcnn_vgg.hdf5
- vgg16_weights_tf_dim_ordering_tf_kernels.h5
- model_vgg_config.pickle
Conclusion: Results of accuracy are pretty good, and the process was fun, educational and interesting.