Robin Cole
Robin Cole
This is currently a feature request. There are 2 parts to your request: - deploying custom classifiers - chaining models, i.e. detect car, then classify it
I expect you can get better % accuracy from the classifier. With obj detection there are all these issues around small object detection which makes them less accurate. Also training...
I think it is a good suggestion, but also a significant breaking change if the current approach is dropped
Prometheus is basically a db? Can it be brought up via a docker compose? Once we have decided to work on this we can evaluate the tradeoffs of the different...
Note to self to also checkout Redis time series, I remember seeing a nice integration with an object detection system https://redislabs.com/blog/my-other-stack-is-redisedge/
My google searches keep bringing me back to https://datasette.io/ which allows building a data explorer app with sqlite backend. A variety of plugins add useful functionality such as previewing thumbnails...
I think this could be as simple as having a process which is monitoring the redis queue, and doing something like https://redislabs.com/blog/how-to-manage-real-time-iot-sensor-data-in-redis/
I believe `MODE=High` is not supported on Jetson, remove this
This is a very relevant repo, focussed on person reId https://github.com/mikel-brostrom/Yolov5_DeepSort_Pytorch
Not all scripts have been updated with the required `if __name__ == '__main__': `