compressai_vision.evaluators#

class compressai_vision.evaluators.BaseEvaluator(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria=None)[source]#
digest(gt, pred)[source]#
static get_coco_eval_info_name(name)[source]#
static get_jde_eval_info_name(name)[source]#
reset()[source]#
results(save_path: str | None = None)[source]#
set_annotation_info(dataset)[source]#
training: bool#
write_results(out, path: str | None = None)[source]#
class compressai_vision.evaluators.COCOEVal(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria='AP')[source]#
digest(gt, pred)[source]#
reset()[source]#
results(save_path: str | None = None)[source]#
training: bool#
class compressai_vision.evaluators.MOT_HiEve_Eval(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria='MOTA')[source]#

A Multiple Object Tracking Evaluator for HiEve

This class evaluates MOT performance of tracking model such as JDE specifically on HiEve

mot_eval()[source]#
training: bool#
class compressai_vision.evaluators.MOT_JDE_Eval(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria='MOTA')[source]#

A Multiple Object Tracking Evaluator

This class evaluates MOT performance of tracking model such as JDE in compressai-vision. BaseEvaluator is inherited to interface with pipeline architecture in compressai-vision

Functions below in this class refers to

The class Evaluator inin Towards-Realtime-MOT/utils/evaluation.py at <Zhongdao/Towards-Realtime-MOT> <Zhongdao/Towards-Realtime-MOT>

Full license statement can be found at <Zhongdao/Towards-Realtime-MOT>

digest(gt, pred)[source]#
static digest_summary(summary)[source]#
mot_eval()[source]#
reset()[source]#
results(save_path: str | None = None)[source]#
training: bool#
class compressai_vision.evaluators.MOT_TVD_Eval(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria='MOTA')[source]#

A Multiple Object Tracking Evaluator for TVD

This class evaluates MOT performance of tracking model such as JDE specifically on TVD

mot_eval()[source]#
training: bool#
class compressai_vision.evaluators.OpenImagesChallengeEval(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria='AP50')[source]#
digest(gt, pred)[source]#
reset()[source]#
results(save_path: str | None = None)[source]#
training: bool#
class compressai_vision.evaluators.VisualQualityEval(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria='psnr')[source]#
static compute_msssim(a, b)[source]#
static compute_psnr(a, b)[source]#
digest(gt, pred)[source]#
reset()[source]#
results(save_path: str | None = None)[source]#
training: bool#
write_results(path: str | None = None)[source]#
class compressai_vision.evaluators.YOLOEval(datacatalog_name, dataset_name, dataset, output_dir='./vision_output/', criteria='AP50')[source]#
training: bool#