Semantic search for e-commerce

Ranking products for user queries

1_DSlvk2PUeiTNTjGbVPb9sw.jpeg

Project description

Visual search engine for e-commerce. The task was to find the most similar item to users' photos.

 

Keypoint approach research using opencv, pytorch, matplotlib and seaborn. Picture vectorization model preparation using Acrcface loss in pytorch. Training process visualization using Visdom library. Silverware image data crawling using urllib. Application deployment on Google Cloud Platform in Docker container.

Technologies

python; numpy, pandas, matplotlib, seaborn, nltk, spacy, scikit-learn, keras; Git, GitHub;