some problem that delopy on device, such as snapdragon snpe

cvg / LightGlue

LightGlue: Local Feature Matching at Light Speed (ICCV 2023)

Apache License 2.0

3.32k stars 321 forks source link

some problem that delopy on device, such as snapdragon snpe #118

Open guker opened 6 months ago

guker commented 6 months ago

I want to deploy lightglue on the mobile device(snapdragon snpe), but snapdragon snpe requires a fixed dimension of input. How can the lightgule ensure effectiveness?

Phil26AT commented 5 months ago

In this case you cannot use point-pruning (disable it by setting width_confidence=-1). Then, you need to ensure that the number of features is of fixed length. You can achieve this by adjusting the extractor configs. For Superpoint, set max_num_keypoints=2048 and detection_threshold=0.