Share your work here ✅

ShopNet: A Neural Network for Product Images (http://www.shopnet.ai/)


Trained a CNN with the amazing techniques mentioned throughout the course and deployed it as a web app. Currently at ~95% accuracy but with room for improvement.

The stats:

  • Training data: 26 classes, approx. 400 samples each, batch size of 16 (see current list of classes)
  • CNN learner with resnet50: cnn_learner(data, arch, metrics=accuracy, wd=0.1, ps=0.01)
  • weight decay: wd=0.1
  • dropout: ps=0.01
  • 3 cycles, 5 epochs each
  • Accuracy: 94.6%
  • Other fine-tuning techniques:
    • data augmentation (rotate, zoom, lighting, warp)
    • progressive resizing from 224 to 352
    • adjusting learning rates
2 Likes