Examples¶

Example predictions from the COCO val set. These are the first images with more than five person annotations that were shared with CC-BY-2.0. These examples are not cherry-picked.

All examples are predicted with the fast model shufflenetv2k16.
For more accurate results, try --checkpoint=shufflenetv2k30.

The examples below are generated with the shown command which is documented in {doc}predict_cli.

In [ ]:

!python -m openpifpaf.predict coco/000000188465.jpg --image-output --long-edge=1025

predicted
Image credit: Photo by Brian Jelonek shared under [CC-BY-2.0].

In [ ]:

!python -m openpifpaf.predict coco/000000114907.jpg --image-output

predicted
Image credit: Photo by Fort Rucker shared under [CC-BY-2.0].

In [ ]:

!python -m openpifpaf.predict coco/000000147740.jpg --image-output

predicted
Image credit: Photo by CCFoodTravel.com shared under [CC-BY-2.0].

In [ ]:

!python -m openpifpaf.predict coco/000000541055.jpg --image-output

predicted
Image credit: Photo by Jarrod Doll shared under [CC-BY-2.0].

Debug Outputs¶

In [ ]:

!python -m openpifpaf.predict coco/000000541055_cropped.jpeg --image-output --debug-images --debug-indices cif:5 cifhr:5 caf:5 --save-all=coco_debug_example --image-height=6 --image-min-dpi=200 --show-file-extension=jpeg

CIF_c CIF_r CIF_hr Image credit: Photo by Jarrod Doll shared under [CC-BY-2.0].

CAF_c CAF_r Image credit: Photo by Jarrod Doll shared under [CC-BY-2.0].

In [ ]: