Open miquel-espinosa opened 3 months ago
@shenqq377 Any thoughts? It should be detecting more instances (like in your provided notebook)
@shenqq377 Any thoughts? It should be detecting more instances (like in your provided notebook) I would like to suggest you double-check the proposals generated by SAM, and the threshold score in feature matching. For the visualization, the NMS threshold also affect the visualization of detection.
Hi.
I am trying to reproduce the results shown in the last image in the notebooks provided. It seems that the results I am obtaining when running the jupyter notebooks are different. I am using the same models and the InsDet full dataset (~6.8G).
What could be the reason of output mismatch? Thanks in advance,
(output obtained with the RankSelect notebook)