This is a fork of Yolo-v3 and Yolo-v2 for Windows and Linux by AlexeyAB for creating a custom model for My MTG card detection project.
Uploading all the progresses on the model training for the last few days.
First batch of model training is completed, where I used ~40,000 generated images of MTG cards laid out in one of the pre-defined pattern.
After 5000 training epochs, the model got 88% validation accuracy on the generated test set.
However, there are some blind spots on the model, notably:
Example of bad detections:
The second and third problems should easily be solved by further augmenting the dataset with random lighting and image skew. I'll have to think more about the first problem, though.
Added several image augmentation techniques to apply to the training set: noise, dropout, light variation, and glaring:
Currently trying to generate enough images to start model training. Hopefully this helps.
Recompiled darknet with OpenCV and CUDNN installed, and recalculated anchors.
I've ran a quick training with tiny_yolo configuration with new training data, and Voila! The model performs significantly better than the last iteration, even against some hard images with glaring & skew! The first prediction model can't detect anything from these new test images, so this is a huge improvement to the model :)
The video demo can be found here: https://www.youtube.com/watch?v=kFE_k-mWo2A&feature=youtu.be
I've been training a new model with a full YOLOv3 configuration (previous one used Tiny YOLOv3), and it's been taking a lot more resources:
The author of darknet did mention that full network will take significantly more training effort, so I'll just have to wait. At this rate, it should reach 50k epoch in about a week :/
