Any progress lately?
Any updates within the last couple of years? Whats the status of sharing the training data? I could help tag/update this.
The training data can be found here: https://github.com/BeeAlarmed/BeeDataset
And no, there is no progress in the last years. Reason is, that I have no access to bees anymore. Hopefullly, I'll have my own bees in the next year(s), then I'll follow up on this. But most likely with a jetson orin and a new approach which is less dependant on the input domain (image resolution, lighting, camera position, etc.). The current solution is a mix of a classical approach (much faster) and AI. The classical, rule-based, parts were specialized heavily on my input data and cannot not be adapted easily to other data, unfortunately. That's why I would start it from scratch witch a more generic process. But at that time, the Jetson Nano couldn't handle it, due to its low performance.