RGCN_T: tracking by a r-gcn framework

MOT17-01-DPM


Benchmark:

MOT17 |

Short name:

RGCN_T

Detector:

Public

Description:

we present a unified graph optimization framework to solve the data as- sociation problem in multiple object tracking. The proposed framework build graphs from detection nodes and formulate the graph optimization to a learnable link prediction problem. Specifically, we first use a relational graph convolution net- work (R-GCN) to learn features of both detection nodes and their linked edges. To exploit various edge information, we then use a message passing network (MPN) to aggregate dif- ferent types of edge representations and propagate them on the graph iteratively. Finally, we build an end-to-end learn- able link prediction model by using the leaned edge features to predict if two nodes should be linked. Experimental re- sults on three benchmark datasets demonstrate the effective- ness of the proposed framework, outperforming state-of-the- art methods in multi-object tracking.

Reference:

Last submitted:

October 02, 2020 (6 months ago)

Published:

October 12, 2020 at 08:40:10 CET

Submissions:

3

Project page / code:

n/a

Open source:

No

Hardware:

3 GHZ

Runtime:

59.2 Hz

Benchmark performance:

Sequence MOTA IDF1 HOTA MOTP MT ML FP FN Rcll Prcn AssA DetA AssRe AssPr DetRe DetPr LocA FAF ID Sw. Frag
MOT1763.966.153.079.4795 (33.8)655 (27.8)22,565179,56868.294.554.252.060.973.455.777.282.11.31,774 (26.0)4,182 (61.3)

Detailed performance:

Sequence MOTA IDF1 HOTA MOTP MT ML FP FN Rcll Prcn AssA DetA AssRe AssPr DetRe DetPr LocA FAF ID Sw. Frag
MOT17-01-DPM50.363.549.476.9811523,15451.198.464.138.170.776.539.776.580.20.1213
MOT17-01-FRCNN58.159.048.581.51053832,28964.591.648.249.253.875.553.776.283.60.931108
MOT17-01-SDP49.159.045.876.29103242,95354.291.553.539.361.570.842.471.679.70.7618
MOT17-03-DPM89.382.168.382.512703,2217,76692.696.863.673.570.077.678.481.984.62.1203719
MOT17-03-FRCNN70.169.153.978.462171,94929,29772.097.553.354.858.276.357.878.381.41.396176
MOT17-03-SDP75.072.156.477.880153,14722,94778.196.354.558.760.273.662.677.280.92.1119232
MOT17-06-DPM62.965.552.280.776482893,93466.696.452.552.058.176.654.979.683.20.2151271
MOT17-06-FRCNN57.058.047.578.985621,4323,58169.685.145.450.068.753.357.370.081.31.256121
MOT17-06-SDP57.759.047.878.995641,4683,46770.685.045.250.770.452.658.170.081.31.249123
MOT17-07-DPM45.350.739.678.09176048,57549.293.241.737.944.871.140.075.780.81.262117
MOT17-07-FRCNN64.564.250.380.22231,0504,83271.492.047.653.351.274.058.575.482.62.1122426
MOT17-07-SDP47.452.040.877.411149797,84653.690.241.640.346.167.643.473.180.22.069148
MOT17-08-DPM31.941.937.981.4163866313,66935.391.850.628.459.367.229.877.483.41.14656
MOT17-08-FRCNN38.040.534.981.1182832612,60440.396.338.132.442.372.033.579.983.10.5157399
MOT17-08-SDP33.542.137.580.9183868713,32836.991.947.629.657.765.231.077.383.01.14353
MOT17-12-DPM46.556.446.881.620413674,25750.992.354.640.262.171.742.877.783.50.41428
MOT17-12-FRCNN56.768.652.780.127205943,13563.890.358.547.863.677.152.474.282.70.724178
MOT17-12-SDP47.055.946.481.219433724,20751.592.353.340.462.369.243.277.483.50.41428
MOT17-14-DPM49.562.344.074.743451,8037,22660.986.247.341.551.870.646.866.278.62.4300572
MOT17-14-FRCNN34.646.734.775.319711,39910,58942.784.939.630.646.860.933.566.678.41.997183
MOT17-14-SDP37.948.336.075.421651,4569,91246.485.539.233.346.160.436.567.378.51.9113213

Raw data: