A Fashion Image Search Engine. Returns images of apparels similar to the input image along with a short description. The Search Engine is created using transfer learning techniques. Pre-Trained CNN architectures (with ImageNet weights) are used and evaluated. Following 3 architectures have been used,
- ResNet 50
- VGG 16
- Xception
Euclidean distance has been used as the similarity metric. VGG 16 has been used as the final model for deployment. Below attached video and images demonstrate the use cases of the Women Fashion Image Search Engine. You can also checkout the video demo by clicking on the link attached below.
FashionImageSearchEngine_Demo_1.mp4
<img width="1512" alt="Screenshot 2022-02-21 at 7 54 27 AM" 
