manu-chauhan / visionX

Deep Vision based repository for many top Computer Vision models for ease of use and experiments. To be enhanced with new approaches with time.

VisionX

PyTorch based repository for structured and modularised code for various Computer Vision models, for results/metrics, ease of use and out of the box experimentation : work in progress

---

How to interpret the repository:

1. Models dir has all CV models, custom models are under custom.py.

2. Utils dir contains utility functions and components.

3. Extra resources like images or gifs would go under resources .

4. Notebooks dir will hold all Jupyter Notebooks for experiments and/or final runs for presenting results.

---

Interesting results/experiments:

• 89K and 95K parameter models on Cifar10 with 85%+ validation accuracy - uses augmentation + groups param
• 93% validation accuracy on Cifar10 with Custom ResNet model - uses augmentation + One Cycle LR