CarDamageScore

Inspiration

Often we miss out good deals in any selling platform due to limited time, an automated system which can help to provide health score could significantly improve users experience.

What it does

We provide a damage score based on car visual inspection

How we built it

We used deep learning algorithm to segment the damaged surface on the car body and calculate the damage area. The algorithm provides a score based on the following metric 100 - (damage_area/total_area)

Challenges we ran into

• Traning/Fine Tuning the model
• CORS issue after deploying in GCP
• HTTPS certificate issue

Accomplishments that we're proud of

Deployed a custom trained model as a REST API in GCP

What we learned

What's next for CarDamageScore

• We will improvise the current metric to calculate score and average damage score over all the pictures

Built With

• computer-vision
• deep-learning
• django
• python
• pytorch
• react
• rest-api