Object Detection with DETR

Inspiration

To provide a more convenient way for people to identify things they don't know.

What it does

Given an image, the DETR model can predict multiple instances of object with bounding boxes and classes. In an ideal world, this model would help us identify objects through the AR experience.

How we built it

We followed a guideline provided by researchers at Facebook.

Accomplishments that we're proud of

We successfully implemented a basic model.

What we learned

We learned that compared with other models such as R-CNN, DETR is more efficient in data preprocessing and prediction post-processing.

What's next for Object Detection with DETR

For next, we plan to combine our model with AR to improve the user experience.

Built With

• python