DPM: Deformable Part-based Model

MOT17-01


Benchmark:

Short name:

DPM

Description:

At a high level DPM can be characterized by the combination of

* Strong low-level features based on histograms of oriented gradients (HOG)
* Efficient matching algorithms for deformable part-based models (pictorial structures)
* Discriminative learning with latent variables (latent SVM)

Reference:

P. Felzenszwalb, R. Girshick, D. McAllester, D. Ramanan. Object Detection with Discriminatively Trained Part Based Models. In TPAMI, 2010.

Last submitted:

February 14, 2017 (3 years ago)

Published:

February 14, 2017 at 06:22:59 CET

Submissions:

3

Open source:

Yes

Hardware:

3GHz

Runtime:

19.7 Hz

Benchmark performance:

Sequence AP MODA MODP FAF TP FP FN Recall Precision F1
MOT17Det0.6131.275.87.178,00742,30836,55768.164.866.4

Detailed performance:

Sequence AP MODA MODP FAF TP FP FN Recall Precision F1
MOT17-010.4320.371.92.72,0771,2312,08949.962.855.6
MOT17-030.6934.476.117.749,93826,47618,18473.365.469.1
MOT17-060.6246.373.81.45,7461,6933,00365.777.271.0
MOT17-070.6632.773.67.16,5823,5582,67871.164.967.9
MOT17-080.7927.380.05.34,8783,30786784.959.670.0
MOT17-120.6214.377.03.33,7132,9411,68168.855.861.6
MOT17-140.3415.074.34.15,0733,1028,05538.662.147.6

Raw data: