search

ucbdrive / few-shot-object-detection

Implementations of few-shot object detection benchmarks
Apache License 2.0
1.1k stars 223 forks source link

about novel files？ #13

Closed dmuqlzhang closed 4 years ago

dmuqlzhang commented 4 years ago

image faster_rcnn_R_101_FPN_ft_novel1_1shot.yaml file is not exist？ can you share it？ image

thomasehuang commented 4 years ago

Hello,

We do not actually do this step for PASCAL VOC, so that's why the config files are not provided. Instead, we just randomly initialize the novel weights. I will copy and paste the config file you highlighted anyways in case you still want to try it out. You can also look at the COCO and LVIS config files to see how they are used.

_BASE_: "../Base-RCNN-FPN.yaml"
MODEL:
  WEIGHTS: "checkpoints/voc/faster_rcnn/faster_rcnn_R_101_FPN_base1/model_reset_remove.pth"
  MASK_ON: False
  RESNETS:
    DEPTH: 101
  ROI_HEADS:
    NUM_CLASSES: 5
    FREEZE_FEAT: True
  BACKBONE:
    FREEZE: True
  PROPOSAL_GENERATOR:
    FREEZE: True
INPUT:
  MIN_SIZE_TRAIN: (480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800)
  MIN_SIZE_TEST: 800
DATASETS:
  TRAIN: ("voc_2007_trainval_novel1_1shot", )
  TEST: ('voc_2007_test_novel1',)
SOLVER:
  IMS_PER_BATCH: 16
  BASE_LR: 0.01
  STEPS: (500, )
  MAX_ITER: 1000
  CHECKPOINT_PERIOD: 1000
  WARMUP_ITERS: 0
OUTPUT_DIR: "checkpoints/voc/faster_rcnn/faster_rcnn_R_101_FPN_ft_novel1_1shot"

Hope that helps! Let me know if you run into any issues with this.

dmuqlzhang commented 4 years ago

Hello,

We do not actually do this step for PASCAL VOC, so that's why the config files are not provided. Instead, we just randomly initialize the novel weights. I will copy and paste the config file you highlighted anyways in case you still want to try it out. You can also look at the COCO and LVIS config files to see how they are used.

_BASE_: "../Base-RCNN-FPN.yaml"
MODEL:
  WEIGHTS: "checkpoints/voc/faster_rcnn/faster_rcnn_R_101_FPN_base1/model_reset_remove.pth"
  MASK_ON: False
  RESNETS:
    DEPTH: 101
  ROI_HEADS:
    NUM_CLASSES: 5
    FREEZE_FEAT: True
  BACKBONE:
    FREEZE: True
  PROPOSAL_GENERATOR:
    FREEZE: True
INPUT:
  MIN_SIZE_TRAIN: (480, 512, 544, 576, 608, 640, 672, 704, 736, 768, 800)
  MIN_SIZE_TEST: 800
DATASETS:
  TRAIN: ("voc_2007_trainval_novel1_1shot", )
  TEST: ('voc_2007_test_novel1',)
SOLVER:
  IMS_PER_BATCH: 16
  BASE_LR: 0.01
  STEPS: (500, )
  MAX_ITER: 1000
  CHECKPOINT_PERIOD: 1000
  WARMUP_ITERS: 0
OUTPUT_DIR: "checkpoints/voc/faster_rcnn/faster_rcnn_R_101_FPN_ft_novel1_1shot"

Hope that helps! Let me know if you run into any issues with this.

thanks for your help， I will try to run.