Open namikosaito opened 1 month ago
The aggregated.pkl
file contains the predicted object poses, while the criterion.pkl
file includes various evaluation metrics. We separate the inference and evaluation processes to avoid re-running lengthy inference on large datasets, which enhances efficiency.
For visualizations, you can use the data in aggregated.pkl
. Additionally, we are developing a version integrated with SAM, which will cover the entire workflow from instance segmentation to object pose estimation, including visualization. Please stay tuned for updates to our repository for this upcoming feature.
We would like to ask about the output of the evaluation code.
As written in Github instruction, we run $ bash scripts/eval_estimate.sh
It seems there are no image outputs and we are wondering if we can get some images like your ping-pong racket indicated with pose.
Also, could you please teach us what the pkl files in GenPose2/results/evaluation_results/single directly which are generated after running the bash? There are "aggregated.pkl", "criterion.pkl", and etc and it would be very helpful if you could tell us what they are.