-
**Describe the bug**
`otx train` with Semisupervised train type throws `TypeError: DataLoader.__init__() got an unexpected keyword argument 'unlabeled'`.
**Steps to Reproduce**
```
…
-
During some very large scans, the scan can gradually slow to a crawl. I ran into this issue `use_previous` from the `asset_inventory` module, where after a certain point, the events slowed to the poin…
-
-
**Describe the bug**
I followed the [object detection](https://openvinotoolkit.github.io/training_extensions/latest/guide/tutorials/base/how_to_train/detection.html) example and tried to train YOLO…
-
### Describe the bug
The script was worked perfectly with my old rx-tx module (QCZEK LRS) but i bought a new Betafpv 868 elrs system. I am getting the necessary telemtry datas, but an another ID with…
-
I installed r2Diaphora using PIP3 on my Kali machine and getting errors/exceptions.
1. Exception while trying to read function at 0x%x in sample %s
2. "KeyError: 'instrs'" : It looks like this ke…
-
### Which AdBlocker/DNS cloud service do you use?
AdGuard DNS
### Other
_No response_
### NextDNS users only
- [ ] I have disabled the block page in NextDNS and confirm that the problem still occ…
-
**Describe the bug**
Training goes well, however, in the evaluation phase, OTX loads the weird checkpoint.
In the figure below, OTX loads the 1 epoch's checkpoint although OTX already had …
-
I trained a YOLOX based object detection with very small objects and it seems that the current input size (`416x416`) is not enough to detect the small objects. How is it possible to change the input …
-
No support for mini gimbal on left side (gimblal with up/down page buttons)
X12s ethos V1.0.1