parthkvv / Lane_Detection_IDD_Dataset

Lane detection using CNN with separate embedding and segmentation branches that yield embeddings on each lane pixel. Adopted instance segmentation approach to alleviate the issue of changing lane counts.

Data Preparation

The training samples consist of three components, a binary segmentation label file, a instance segmentation label file and the original image. The binary segmentation uses 255 to represent the lane field and 0 for the rest. The instance use different pixel value to represent different lane field and 0 for the rest.

Image scaling will be done according to config file in similar manner for all images.

Generating tensorflow records file

python tools/make_tusimple_tfrecords.py 

Set the path till "dataset\train_set" in config.py file

"dataset\train_set" directory should consist of following files:
	"train_img" - Folder containing all the ground truth images
	"train_seg_img" - Folder containing all the binary label images

	"val_img" - Folder containing all the ground truth images
	"val_seg_img" - Folder containing all the binary label images

txt_file_gen_SCNN.py
	- generate txt files from image dataset, saved in dataset/train_set/seg_label/list
	  "train_gt.txt"
	  "val_gt.txt"
	  "test_gt.txt"	

Train

• tools/train.py --exp_dir ./experiments/exp0

saved models - \experiments\exp0\

Number of epochs 
- \experiments\exp0\cfg.json
"MAX_EPOCHES": 60

python tools/train_lanenet_tusimple.py 

Test

- \tools\demo_test.py -i E:\Abhishek\Lane_Detection\CULane\parth\SCNN\SCNN_Pytorch-master\demo\demo.jpg -w E:\Abhishek\Lane_Detection\CULane\parth\SCNN\SCNN_Pytorch-master\experiments\exp0\exp0_best.pth


- EVALUATE ON CUSTOM TEST DATASET :
	test_tusimple.py --exp_dir ./experiments/exp0 (keep one category of test data at a time)

Evaluation

- \dataset\Evaluate\
	- Generate csv files with results