Printed circuit boards (PCBs) are the primary component of any electric design and are used in all kinds of fields such as logistics, defense, aeronautics, the medicine industry ,and the automobile industry. With the increasing popularity of consumer electronic products, accurate PCB manufacturing is critical along with production at scale, this calls for accurate automated quality inspection systems that can provide fast and quantitative information of defects and therefore prove to be an asset in the manufacturing process.
We aim to use semantic segmentation to detect inconsistencies in circuit board images by using pre-recorded and labeled circuit imagery of working and non-working PCBs. Semantic segmentation involves assigning a class to every pixel of the image, wherein we treat multiple members of a class as a single entity. Semantic segmentation would let us examine the circuit network imagery and detect inconsistencies in those networks.
This is a DL based project which uses Sementic segimentaion thorough UNET algorithm for detection of the broken joins of faults in a PCB and resulta the exact position in form of bounding area and texual format that weather the given PCB had an error or not.
- Clone the repository to your local enviornment
- Decide weather you want to tarin upon Torch or TensorFlow
- Crate the directory structure required for the image folders
- run
train.py
by specifying the required information - The final model and checkpoints will be saved to your local runtime which can be used for further detection