generated from tlpss/research-template
-
Notifications
You must be signed in to change notification settings - Fork 1
Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
- Loading branch information
Showing
5 changed files
with
376 additions
and
0 deletions.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
1 change: 1 addition & 0 deletions
1
state-estimation/state_estimation/keypoint_detection/akd/.gitignore
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1 @@ | ||
| **.json |
199 changes: 199 additions & 0 deletions
199
state-estimation/state_estimation/keypoint_detection/explore_akd.ipynb
Large diffs are not rendered by default.
Oops, something went wrong.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
133 changes: 133 additions & 0 deletions
133
state-estimation/state_estimation/keypoint_detection/keypoint_average_distance_metric.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,133 @@ | ||
| import torch | ||
| from keypoint_detection.data.coco_dataset import COCOKeypointsDataset | ||
| from keypoint_detection.models.detector import KeypointDetector | ||
| from keypoint_detection.utils.heatmap import get_keypoints_from_heatmap_batch_maxpool | ||
| from keypoint_detection.utils.load_checkpoints import get_model_from_wandb_checkpoint | ||
| from torch.utils.data import DataLoader | ||
| from tqdm import tqdm | ||
|
|
||
|
|
||
| def calculate_average_error_for_dataset( | ||
| model: KeypointDetector, dataset_json_path, channel_config: list[list[str]], detect_only_visible_keypoints | ||
| ): | ||
| dataset = COCOKeypointsDataset( | ||
| dataset_json_path, | ||
| keypoint_channel_configuration=channel_config, | ||
| detect_only_visible_keypoints=detect_only_visible_keypoints, | ||
| ) | ||
| dataloader = DataLoader(dataset, batch_size=1, shuffle=False, num_workers=1) | ||
|
|
||
| model.eval() | ||
| model.cuda() | ||
|
|
||
| errors = [[] for _ in range(len(channel_config))] | ||
| for image, keypoints in tqdm(dataloader): | ||
| image = image.cuda() | ||
| heatmaps = model(image) | ||
| keypoints = keypoints | ||
| predicted_keypoints = get_keypoints_from_heatmap_batch_maxpool(heatmaps, max_keypoints=1)[0] | ||
| for i in range(len(channel_config)): | ||
| if len(predicted_keypoints[i]) == 0: | ||
| print("no keypoints found") | ||
| continue | ||
| if len(keypoints[i]) == 0: | ||
| # print("no GT keypoints found") | ||
| continue | ||
| kp = torch.tensor(predicted_keypoints[i][0], dtype=torch.float32) | ||
| gt_kp = torch.tensor(keypoints[i][0], dtype=torch.float32) | ||
| l2_error = torch.norm(kp - gt_kp) | ||
| errors[i].append(l2_error.item()) | ||
|
|
||
| average_errors = [sum(errors[i]) / len(errors[i]) for i in range(len(channel_config))] | ||
| for i in range(len(channel_config)): | ||
| print(f"Average error for channel {channel_config[i]}: {average_errors[i]}") | ||
| print(f"Average error: {sum(average_errors)/len(average_errors)}") | ||
|
|
||
| mae_dict = {} | ||
| full_dict = {} | ||
| for i in range(len(channel_config)): | ||
| channel_name = "" + "-".join(channel_config[i]) | ||
| mae_dict[channel_name] = average_errors[i] | ||
| full_dict[channel_name] = errors[i] | ||
| mae_dict["average"] = sum(average_errors) / len(average_errors) | ||
|
|
||
| return mae_dict, full_dict | ||
|
|
||
|
|
||
| if __name__ == "__main__": | ||
| from state_estimation.keypoint_detection.common import ( | ||
| SHORTS_CHANNEL_CONFIG, | ||
| TOWEL_CHANNEL_CONFIG, | ||
| TSHIRT_CHANNEL_CONFIG, | ||
| data_dir, | ||
| ) | ||
| from state_estimation.keypoint_detection.final_checkpoints import ARTIFACT_DICT | ||
| from state_estimation.keypoint_detection.real_baselines import ( | ||
| ARTF_SHORTS_TEST_PATH, | ||
| ARTF_TOWEL_TEST_PATH, | ||
| ARTF_TSHIRT_TEST_PATH, | ||
| ) | ||
|
|
||
| error_dict = {} | ||
| for key, value in ARTIFACT_DICT.items(): | ||
| if "tshirt" in key: | ||
| wandb_checkpoint = value | ||
| if "sim" in key: | ||
| dataset = ( | ||
| data_dir | ||
| / "synthetic-data" | ||
| / "TSHIRT" | ||
| / "single-layer-random-material-10K" | ||
| / "annotations_val.json" | ||
| ) | ||
| else: | ||
| dataset = ARTF_TSHIRT_TEST_PATH | ||
| keypoints = TSHIRT_CHANNEL_CONFIG.split(":") | ||
| keypoints = [channel.split(",") for channel in keypoints] | ||
| elif "towel" in key: | ||
| wandb_checkpoint = value | ||
| if "sim" in key: | ||
| dataset = ( | ||
| data_dir / "synthetic-data" / "TOWEL" / "single-layer-random-material-10K" / "annotations_val.json" | ||
| ) | ||
| else: | ||
| dataset = ARTF_TOWEL_TEST_PATH | ||
| keypoints = TOWEL_CHANNEL_CONFIG.split(":") | ||
| keypoints = [channel.split(",") for channel in keypoints] | ||
| elif "shorts" in key: | ||
| wandb_checkpoint = value | ||
| if "sim" in key: | ||
| dataset = ( | ||
| data_dir | ||
| / "synthetic-data" | ||
| / "SHORTS" | ||
| / "single-layer-random-material-10K" | ||
| / "annotations_val.json" | ||
| ) | ||
| else: | ||
| dataset = ARTF_SHORTS_TEST_PATH | ||
| keypoints = SHORTS_CHANNEL_CONFIG.split(":") | ||
| keypoints = [channel.split(",") for channel in keypoints] | ||
| else: | ||
| raise ValueError("Unknown artifact key") | ||
|
|
||
| print(f"Calculating average error for {key}") | ||
| print(f"dataset = {dataset}") | ||
| print(f"keypoints = {keypoints}") | ||
|
|
||
| model = get_model_from_wandb_checkpoint(wandb_checkpoint).cuda() | ||
| avg_errors, d = calculate_average_error_for_dataset( | ||
| model, dataset, keypoints, detect_only_visible_keypoints=True | ||
| ) | ||
| error_dict[key] = avg_errors | ||
|
|
||
| # save dict as json | ||
| import json | ||
| import pathlib | ||
|
|
||
| file_path = pathlib.Path(__file__).parent | ||
| # with open(file_path / "average_keypoint_distances.json", "w") as f: | ||
| # json.dump(error_dict, f, indent=4) | ||
|
|
||
| with open(file_path / "akd" / f"{key}.json", "w") as f: | ||
| json.dump(d, f, indent=4) |