By Yanwei Pang†, Tiancai Wang†, Rao Muhammad Anwer, Fahad Shahbaz Khan, Ling Shao

† denotes equal contribution

Single-stage object detectors have recently gained popularity due to their combined advantage of high detection accuracy and real-time speed. However, while promising results have been achieved by these detectors on standard-sized objects, their performance on small objects is far from satisfactory. To detect very small/large objects, classical pyramid representation can be exploited, where an image pyramid is used to build a feature pyramid (featurized image pyramid), enabling detection across a range of scales. Existing single-stage detectors avoid such a featurized image pyramid representation due to its memory and time complexity. In this paper, we introduce a light-weight architecture to efficiently produce featurized image pyramid in a single-stage detection framework.

• Clone this repository. This repository is mainly based on ssd.pytorch and RFBNet.

    LFIP_ROOT=/path/to/clone/LFIP
    git clone https://github.com/vaesl/LFIP $LFIP_ROOT

• The code was tested on Ubuntu 16.04, with Anaconda Python 3.5/6 and PyTorch v0.3.1. NVIDIA GPUs are needed for testing. After install Anaconda, create a new conda environment, activate the environment and install pytorch0.3.1.

    conda create -n LFIP python=3.5
    source activate LFIP
    conda install pytorch=0.3.1 torchvision -c pytorch

• Install opencv.

    conda install opencv

• Compile both COCOAPI and NMS:

    cd $LFIP_ROOT/
    ./make.sh

To evaluate the performance reported in the paper, Pascal VOC and COCO dataset as well as our trained models need to be downloaded.

• Directly download the images and annotations from the VOC website and put them into $LFIP_ROOT/data/VOCdevkit/.

• Create the VOCdevkit folder and make the data(or create symlinks) folder like:

  ${$LFIP_ROOT}
  |-- data
  `-- |-- VOCdevkit
      `-- |-- VOC2007
          |   |-- annotations
          |   |-- ImageSets
          |   |-- JPEGImages
          |-- VOC2012
          |   |-- annotations
          |   |-- ImageSets
          |   |-- JPEGImages
          |-- results
  

• Download the images and annotation files from coco website coco website.

• Place the data (or create symlinks) to make the data folder like:

  ${$LFIP_ROOT}
  |-- data
  `-- |-- coco
      `-- |-- annotations
          |   |-- instances_train2014.json
          |   |-- instances_val2014.json
          |   |-- image_info_test-dev2015.json
          `-- images
          |   |-- train2014
          |   |-- val2014
          |   |-- test2015
          `-- cache
  

Please access to Google Driver or BaiduYun Driver to obtain our trained models for PASCAL VOC and COCO and put the models into corresponding directory(e.g. '~/weights/COCO/LFIP_COCO_300/'). Note that the access code for the BaiduYun Driver is jay3 and for the time being we only release models with 300*300 input size.

To check the performance reported in the paper:

python test_LFIP.py -d VOC -s 300 --trained_model /path/to/model/weights

where '-d' denotes datasets, VOC or COCO and '-s' represents image size, 300 or 512.

Please cite our paper in your publications if it helps your research:

@article{Pang2019LFIP,
    title = {Efficient Featurized Image Pyramid Network for Single Shot Detection},
    author = {Yanwei Pang, Tiancai Wang, Rao Muhammad Anwer, Fahad Shahbaz Khan, Ling Shao},
    booktitle = {CVPR},
    year = {2019}
}