Shojin Tam comments

Results 13 comments of


                                            Shojin Tam

RandomResize during inference?

I tried feeding images a few times and found that the images were resized while keeping the original ratio. The short side of the image will be resized to 800...

ValueError: mutable default <class 'hydra.conf.JobConf.JobConfig.OverrideDirname'> for field override_dirname is not allowed: use default_factory

Try using python 3.8 instead https://github.com/geekyutao/Inpaint-Anything/issues/125

issue with coco evaulate

See https://github.com/cocodataset/cocoapi/issues/421

train bug

Sorry. This is my first pull request, so I was a bit clumsy. Please feel free to look at my commit.

Exception in triangulation: 'pycolmap._core.Rigid3d' object has no attribute 'essential_matrix'

Downgrade pycolmap to 3.10.0

'numpy.float64' error

See https://github.com/cocodataset/cocoapi/issues/421

Error: tfrecord/helen.record; No such file or directory

Try this instead https://github.com/google-ai-edge/mediapipe/blob/v0.8.10/mediapipe/modules/face_landmark/face_landmark.tflite

[Solution] ValueError from transformers due to torch.load safety check

I upgraded PyTorch to 2.6, and the code ran without issues.

Dinov2 encoder returns matrix with none values

I have changed this line https://github.com/Parskatt/RoMa/blob/a1494f87ab85485ad8a92a981876ca837e2334e4/demo/demo_match.py#L28 to ```python roma_model = roma_outdoor(device=device, coarse_res=560, upsample_res=(864, 1152), amp_dtype=torch.bfloat16) ``` However, the problem is not resolved.

Dinov2 encoder returns matrix with none values

I'm just sending in the images in the assets directory (toronto_A.jpg and toronto_B.jpg). ```python # toronto_A.jpg np.array(im1) array([[[130, 165, 194], [130, 165, 194], [129, 164, 195], ..., [ 29, 51,...