VISUALIZE.md

Visualization

This page includes various examples on visualization.

Single input

For tasks like classification and pose estimation, which need only one image as input, you should modify your test image in imgs/single.json.

Pedestrian Attribute Recognition

To visualize the result of attribute recognition model,

python3 launch.py --config configs/attr_deepmar/deepmar_rapv2_r50v1.py --task visualize --image imgs/single.json

Pose Estimation

To visualize the result of pose estimation model,

python3 launch.py --config configs/pose_simple_baseline/simple_pose_r50v1.py --task visualize --image imgs/single.json

Dense Human Pose Estimation (DensePose)

To visualize the result of densepose model,

python3 launch.py --config configs/densepose_baseline/dense_pose_r50v1.py --task visualize --image imgs/single.json

Pair of inputs

For person re-identification, you should input two images. The input setting is formatted in imgs/pair.json.

Person re-identification

To visualize the result of re-identification model,

python3 launch.py --config configs/reid_strong_baseline/strong_baseline_market1501_r50v1_xent_tri_cent.py --task visualize --image imgs/pair.json