Open danielbdr opened 8 months ago
tests
cd /home/dbendaya/work/ContinualLearning/tinyYolov3_lava/YOLOsdnn/loihi8YOLO/
import sys
path = ['/home/dbendaya/work/ContinualLearning/tinyYolov3_lava/YOLOsdnn/',
'/home/dbendaya/work/ContinualLearning/prophesee-automotive-dataset-toolbox/']
sys.path.extend(path)
from object_detection.boundingbox.utils import temporal_NMS, storeData, accuracy
from tqdm import tqdm
import numpy as np
T = 10
t_nms = temporal_NMS(16)
accuracy_t_nms = []
for k in tqdm(range(13), desc='loading BDD_on_yolo2hd predictions-accuracy tests on NMS methods:'):
[inputs, targets, bboxes, predictions, counts] = storeData.load('tinyYolo_dump%02d.pkl'%k)
t_nms.reset()
detections_frame = [t_nms(predictions[..., t]) for t in range(T)]
accuracy_t_nms.append(accuracy(detections_frame, bboxes))
print(accuracy_t_nms, np.array(accuracy_t_nms).mean(), sep='\n')
temporal_NMS class extends the static frame NMS considering successive frames. Improves accuracy by 2.5%-3% mAP over the static NMS on DBBK100 dataset
Other utility functions:
Issue Number:
Objective of pull request:
Pull request checklist
Your PR fulfills the following requirements:
flakeheaven lint src/lava tests/
) and (bandit -r src/lava/.
) pass locallypytest
) passes locallyPull request type
Please check your PR type:What is the current behavior?
-
What is the new behavior?
-
Does this introduce a breaking change?
Supplemental information