The project was allotted to our team by Edge Techno company who was in turn working for TVS, the aim was to identify the following defects that might occur in the two wheelers at the time of their manufacturing:
- Tank eccentric
- Tank height
- CBTC flush
- Side panel sticker presence
Need for the project
Manufacturing defects are a major problem for companies. If the product is delivered with these defefcts being left unnoticed during manufacturing, it would result in customer dissatisfaction and the replacement / repair costs will be a burden on the company. If on the other hand the automated system identifies these defects the same may be corrected at the manufacturing time itself.
The estimated cost savings will be a lot in case the company adopts for the automated defect detection. The system will also prove beneficial in protecting the brand value and preserving customer satisfaction.
The project is handled in two stages : defect region identification, defect identification.
The areas where the defects might occur are first identified using deep learning model and then the the region alone is cropped and processed further to identify if the defect is there or not.
We were able to make a GUI based application which succesfully identified. The video demonstration of the same is as follows: