Commit ab41d801 authored by 李聪聪's avatar 李聪聪
Browse files

optimize and add Masks url

parent bcc505b7

README.md

+3 −1
Original line number Diff line number Diff line
@@ -2,7 +2,7 @@ 


Introduction

-----------------

This paper was submitted to ACM MM 2019.

This paper was accepted to ACM MM 2019.



This repository implements DPNet ([Data Priming Network for Automatic Check-Out](/)) using PyTorch 1.0.1 . This implementation is heavily influenced by the project [maskrcnn-benchmark](https://github.com/facebookresearch/maskrcnn-benchmark).
@@ -36,6 +36,8 @@ Using `toolboxes` to extract masks, train [Salient Object Detection](https://git 
and render with [CycleGAN](https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix). Then modify `maskrcnn_benchmark/config/paths_catalog.py` 

to make the paths correct.



The extracted masks can be download here: [Google Drive](https://drive.google.com/drive/folders/1QqPQ0l7nDpIt8SnuUYWaYQU_8e_anHFc?usp=sharing)



## Single GPU training



Most of the configuration files that we provide assume that we are running on 4 GPUs.

configs/e2e_faster_rcnn_R_101_FPN_1x_rpc_finetune.yaml

0 → 100644
+39 −0
Original line number Diff line number Diff line 
MODEL:

  META_ARCHITECTURE: "GeneralizedRCNN"

  WEIGHT: "CHANGE/ME"

  BACKBONE:

    CONV_BODY: "R-101-FPN"

  RESNETS:

    BACKBONE_OUT_CHANNELS: 256

  RPN:

    USE_FPN: True

    ANCHOR_STRIDE: (4, 8, 16, 32, 64)

    PRE_NMS_TOP_N_TRAIN: 2000

    PRE_NMS_TOP_N_TEST: 1000

    POST_NMS_TOP_N_TEST: 1000

    FPN_POST_NMS_TOP_N_TEST: 1000

  ROI_HEADS:

    USE_FPN: True

  ROI_BOX_HEAD:

    POOLER_RESOLUTION: 7

    POOLER_SCALES: (0.25, 0.125, 0.0625, 0.03125)

    POOLER_SAMPLING_RATIO: 2

    FEATURE_EXTRACTOR: "FPN2MLPFeatureExtractor"

    PREDICTOR: "FPNPredictor"

    NUM_CLASSES: 201

DATASETS:

  TRAIN: ("rpc_2019_pseudo",)

  TEST: ("rpc_2019_val",)

DATALOADER:

  SIZE_DIVISIBILITY: 32

  ASPECT_RATIO_GROUPING: False

SOLVER:

  BASE_LR: 0.001

  WEIGHT_DECAY: 0.0001

  STEPS: (50000, 60000)

  MAX_ITER: 70000

  IMS_PER_BATCH: 8

TEST:

  IMS_PER_BATCH: 4



OUTPUT_DIR: 'outputs_rpc_2019_finetune'
 
 No newline at end of file

configs/e2e_faster_rcnn_R_101_FPN_1x_rpc_render.yaml

+1 −0
Original line number Diff line number Diff line
@@ -26,6 +26,7 @@ DATASETS: 
  TEST: ("rpc_2019_val",)

DATALOADER:

  SIZE_DIVISIBILITY: 32

  ASPECT_RATIO_GROUPING: False

SOLVER:

  BASE_LR: 0.01

  WEIGHT_DECAY: 0.0001

configs/e2e_faster_rcnn_R_101_FPN_1x_rpc_render_density_map.yaml

+1 −0
Original line number Diff line number Diff line
@@ -27,6 +27,7 @@ DATASETS: 
  TEST: ("rpc_2019_val",)

DATALOADER:

  SIZE_DIVISIBILITY: 32

  ASPECT_RATIO_GROUPING: False

SOLVER:

  BASE_LR: 0.01

  WEIGHT_DECAY: 0.0001

configs/e2e_faster_rcnn_R_101_FPN_1x_rpc_syn.yaml

+1 −0
Original line number Diff line number Diff line
@@ -26,6 +26,7 @@ DATASETS: 
  TEST: ("rpc_2019_val",)

DATALOADER:

  SIZE_DIVISIBILITY: 32

  ASPECT_RATIO_GROUPING: False

SOLVER:

  BASE_LR: 0.01

  WEIGHT_DECAY: 0.0001