search

matterport / Mask_RCNN

Mask R-CNN for object detection and instance segmentation on Keras and TensorFlow
Other
24.75k stars 11.71k forks source link

Mask R-CNN parameters for tiny objects detection #1071

Open Ujang24 opened 6 years ago

Ujang24 commented 6 years ago

Dear @waleedka, thank you for your code. I have been trying around object detection with your code. I was trying to do cells detection which sizes around 12x12 pixel each. Nevertheless, there are still a lot of undetected cells.

I saw there were a lot of parameters inside. I found on config.py. I think that was for detecting large objects. I have tried to configure a lot of them to have a better detection for my tiny cells. I have tried to reduce the into MINI_MASK_SHAPE = (14, 14), increase RPN_NMS_THRESHOLD = 0.9, reduce DETECTION_MIN_CONFIDENCE = 0.1, DETECTION_MAX_INSTANCES = 500, MAX_GT_INSTANCES = 500, etc.

So do you have suggestions what are the parameters to be edited and what are the values? The image was 640x480 px The objects are 12x12 size with more than 300 objects/image.

Thanks

ShibaPrasad commented 6 years ago

I am getting below error

ValueError Traceback (most recent call last) /miniconda/lib/python3.6/site-packages/tensorflow/python/framework/op_def_library.py in _apply_op_helper(self, op_type_name, name, keywords) 509 as_ref=input_arg.is_ref, --> 510 preferred_dtype=default_dtype) 511 except TypeError as err: ** would you please help me !!!!!