lix-byte / deep-fashion-retrieval

Simple image retrival on deep-fashion dataset with pytorch - A course project

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool


Simple image retrieval algorithm on deep-fashion dataset with pytorch



  • Python (Compatible with 2 and 3)
  • Pytorch
  • Torchvision
  • PIL
  • cv2(only for visualizing)

Anaconda is recommended.


  1. Download dataset from DeepFashion: Attribute Prediction
  2. Unzip all files and set DATASET_BASE in
  3. Run

The models will be saved to DATASET_BASE/models.

My model: Download from Google Drive

Deep Feature: ResNet50 - (Linear 1024 to 512) - (Linear 512 to 20), the 512-dim vector is regarded as images' identical features.

Loss: CrossEntropyLoss + TripletMarginLoss * Weight

Color Feature: Get ResNet50 final conv layer output(N * C * 7 * 7), then do avg_pooling on channel dim. Choose the max-N responses and extract the corresponding blocks on avg_pooling map of original image.

Training details: Freeze the conv parameters and train net until a stable accuracy and loss, then set FREEZE to False and train it again.

If you applied the DeepFashion: In-shop Clothes Retrieval, you can set ENABLE_INSHOP_DATASET to True.

Directory structure:

    • Anno
    • Eval
    • img
    • models(generated)
    • in_shop(optional)
      • list_bbox_inshop.txt
      • list_eval_partition.txt
      • img
    • all_feat.list(generated)
    • all_feat.npy(generated)
    • all_color_feat.npy(generated)

Generating feature databases

  • Feature extraction

    • Set DUMPED_MODEL in as trained model
    • Run

    The feature will be saved to DATASET_BASE/all_feat.npy, DATASET_BASE/all_color_feat.npy and DATASET_BASE/all_feat.list.

  • Accelerating querying by clustering

    • Run to train the models, default 50 clusters.

      The model will be saved as DATABASE/models/kmeans.m

Query with a picture

  • Run img_path, for example:

    python img/Sheer_Pleated-Front_Blouse/img_00000005.jpg.

    Set DISTANCE_METRIC in to use different metrics such as cosine, euclidean on deep feature and color feature..

Time cost

  • 2.854 sec for loading model
  • 0.078 sec for loading feature database
  • 0.519 sec for extracting feature of given image
  • 0.122 sec for doing naive query(139,709 features)
  • 0.038 sec for doing query with kmeans(139,709 features)


  • Intel(R) Core(TM) i7-4790K CPU @ 4.00GHz with 32GB RAM
  • GeForce GTX TITAN X with CUDA 7.5
  • Ubuntu 14.04
  • Pytorch 0.2.0_4

Future works

  • Add web support
  • Add more models and fuse them


Simple image retrival on deep-fashion dataset with pytorch - A course project


Language:Python 100.0%