This project uses a Bi-Directional Cascade Network(BDCN) based on the repository https://github.com/pkuCactus/BDCN to detect edges in rock blasting images to obtain the granulometry study of the fragmented material.
- pytorch >= 1.3.1
- numpy >= 1.11.0
- pillow >= 3.3.0
- opencv >= 4.1.2
- sklearn >= 0.21.3
- scipy >= 1.4.1
- imutils >= 0.5.3
- skimage >= 0.15.0
git clone https://github.com/erikperez20/Granulometry-Edge-Detection-BDCN.git
BDCN model for BSDS500 dataset and NYUDv2 datset of RGB and depth are availavble on Baidu Disk.
The link https://pan.baidu.com/s/18PcPQTASHKD1-fb1JTzIaQ
code: j3de
python rock_detect.py
Options
-h, --help show this help message and exit
--images_file IMAGES_FILE
File where images are stored (ex. Data\Imagen_Prueba)
-c, --cuda use --cuda if using in cpu, else nothing
-g GPU, --gpu GPU the gpu id to run net
-m MODEL, --model MODEL
the model to test (defaults to bdcn_pretrained_on_bsds500.pth)
Rock blasting input image taken from mine dataset
Rock detection results after applying BDCN model, morphological transformations and image thresholding.
Rock sizes distribution considering a scale/size = 1, Gaudin Schuhmann rock distribution, Rosin Rammler distribution and Swebrec distribution.