Uyoung Jeong

icon indicating a website link https://uyoung-jeong.github.io/

icon company UNIST icon location PhD course student at UNIST, South Korea

Results 4 comments of Uyoung Jeong

Error during single dataset evaluation

After fixing dataloader_nb=0, I got another error in spec/trainer.py https://github.com/mkocabas/SPEC/blob/d2fe2c264c72c98a5f479fc36f74bdd5f45427da/spec/trainer.py#L524 AttributeError: 'dict' object has no attribute 'append' I added an input argument `ds_name=self.val_ds[0].dataset` to prevent the error.

Reproduction of the numbers in the paper

I got different result. MPJPE: 99.1523, PA-MPJPE: 64.8514, PVE: 912.4414, ACCEL: 31.4988, ACCEL_ERR: 31.2689. Still not the same as the paper.

'with_renderer' argument raises TypeError

I fixed the 485th line of image_base.py as below, and it works. `visualizer = Visualizer(resolution = (512,512,3), result_img_dir=save_dir, renderer_type=args().renderer)`

Reproduced results on COCO val dataset are inferior than official results

This might happen when you train this model with smaller batch size. Training with 2 gpus, each with 20 batch size will produce similar or slightly higher result.