FAIR's research platform for object detection research, implementing popular algorithms like Mask R-CNN and RetinaNet.
26.21k
stars
5.45k
forks
source link
Not all items detected in picture (just not a big part) #1008
Closed
revaldo666 closed 3 years ago
Expected results
I have a list of 250 trained images, like this, (average each image size 1000px X 3000px with ~ 300 polygons in each)![image](https://user-images.githubusercontent.com/2873127/97813067-63c94280-1c8e-11eb-9c7b-ecf17f378beb.png)
but when I'm trying to check the trained model, only part of the evaluation image successfully masked. What I'm doing wrong?
Actual results
This is the image after prediction:![image](https://user-images.githubusercontent.com/2873127/97813152-e3efa800-1c8e-11eb-9e45-5ccb0945a1b7.png)
Not all stones detected.
Detailed steps to reproduce
Trainer configuration:
cfg = get_cfg() cfg.merge_from_file(model_zoo.get_config_file("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml")) cfg.DATASETS.TRAIN = ("balloon_train",) cfg.DATASETS.TEST = () cfg.DATALOADER.NUM_WORKERS = 2 cfg.MODEL.WEIGHTS = model_zoo.get_checkpoint_url("COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml") cfg.SOLVER.IMS_PER_BATCH = 2 cfg.SOLVER.BASE_LR = 0.01 # pick a good LR cfg.SOLVER.WARMUP_ITERS = 300 cfg.SOLVER.CHECKPOINT_PERIOD = 300 cfg.SOLVER.MAX_ITER = 10000
cfg.SOLVER.STEPS = (1000,) cfg.SOLVER.GAMMA = 0.005 cfg.MODEL.ROI_HEADS.POSITIVE_FRACTION = 0.7 cfg.MODEL.ROI_HEADS.IOU_THRESHOLDS = [0.5] cfg.MODEL.ROI_HEADS.BATCH_SIZE_PER_IMAGE = 32
cfg.MODEL.ROI_HEADS.NUM_CLASSES = 1 cfg.TEST.EVAL_PERIOD = 300