- Using deep learning approach, to build a convolution neural network which will give the bounding box of object present in input image.
- The model should work for any flipkart product given in dataset.
- Tried various model architectures
- Simple modular code for easy managemet
- Used Keras - deep learning framework of python
- Each experiment folder contains different CNN model architectures
- Each folder contains
Model.py - Contains the main CNN model architecture **hyperparameter.py **- Contains the respective hyperparameters of CNN model **data_loader.py **- Loads the image dataset into main memory in batches for training and testing output_gen.py - Generates the output and stores it in csv file. plot.py - Can be used to see the output train.py - Where all other utilities are called and model is trained. Trained model is saved per epoch. This is the main file which we run before output gen and ploting.
- As the image dataset was around 14GB, it was not possible to upload it on github.
- As I tried various models, there were too many saved models. As model size is huge I couldn't upload it on github.