Inspiration
Current free APIs for Gemini, Claude and even GPT are not that accurate given an image to LLMs. Our fine tuned model on custom dataset called DeepFashion2.
What it does
It tries to recommend similar products on the market places available like, H&M, Amazon, Zara etc. Also, it predicts with confidence value, how confident model is.
How we built it
We have used YOLOV26n object detection for dataset available called DeepFashion2 (~1,91,000 images) which is used for training, testing and validation.
Challenges we ran into
Training using GPU was problematic as there were lot of images to train.
Accomplishments that we're proud of
Integration with AI and our own training of pretrained model with confidence value is something we are proud to announce. As we tried to use free APIs available for LLMs were not that enough to detect specific outfit that person is wearing.
What we learned
We learned how to utilize our own CUDA GPU for training and use Pytorch framework.
What's next for AI based outfit detection and product recommendation
We can integrate more inventories and scale this product to the next level. There could be multiple vendors and similar outfits that we can add to our inventory to match the best fit and recommend respectively.
Built With
- cnn
- database
- fastapi
- react
- yolo
Log in or sign up for Devpost to join the conversation.