GERMAN TRAFFIC SIGN DETECTION

Detect The German Traffic Sign Using YoloV5

DATA COLLECTION:

Data Collected form Rubixe For Educational Internship

TASK: OBJECT DETECTION

DATA INFORMATION:

Total 4290 traffic sign images are present in 39 classes Each class contain 110 images
Split data with the help of splitfolder library.
88 Images for training and 22 images for validation
Create bounding boxes with the help of label-img tool and makesense.ai website

German Traffic Signs

DATA PREPRATION:

Prepare folder structure that can be accept by YoloV5.
Total 3438 images for training and 857 images for validation present in 39 classes.
Create a bounding boxes with the help of label-img And makesense.ai website according to YoloV5.

STEPS TO USE YOLOV5

Cloning the YoloV5 file from official repository.
Changing the directory of yolov5
Installing the dependencies
Download all versions pre-trained weights

STEPS BEFORE TRAINING CUSTOM DATASET:

Go to yolov5/data/
Open coco128.yaml
Edit the following inside it:

A. Training and Validation file path

B. Number of classes and Class names.

TRAINING YOLOV5 MODEL

Set images size 128 with batch of 8
Train model on 50 epochs
Gives the data file path as well as give pre-trained weights path.

VISUALISE THE TRAINING METRICS WITH THE HELP OF TENSORBOARD

AFTER TRAINING THE MODEL

VALIDATION IMAGES PREDICTION:

PREDICTED IMAGES:

CHALLENGES FACED:

Facing problem to understand the business case.
challenge faced in bounding boxes creation
Assign same no for all classes
Made mistake in yolov5 folder structure
Take lots of time to create bounding boxes

WHAT WE LEARN:

Convert classification task to object detection to improve skill in object detection
Understand the YoloV5 folder structure as well as learn label-img tool.
Learn pytorch library.