ICME2020 Grand Challenge: The PAIR competition ended successfully

2020-03-09

 

The “ICME2020 Grand Challenge-Embedded Deep Learning Object Detection Model Compression Competition for Traffic in Asian Countries” co-sponsored by PAIR Labs and iVS Lab (Directed by Prof. Jiun-In Guo) has successfully concluded on March 6. This competition is mainly aimed at the unique transportation and road conditions in Asia. After two months of qualification competition and four weeks of final competition, the winners and special awards were finally released.(Photo source: Pixabay)

 

Award Winners

  • Champion: USTC-NELSLIP
  • First Runner-up: BUPT_MCPRL
  • Second Runner-up: DD_VISION

 

Special Awards

  • Best accuracy award: BUPT_MCPRL
  • Best bicycle detection award: IBDO-AIOT
  • Best scooter detection award: Deep Learner

 

Final Evaluation Result

 

group Name Accuracy % Model Size

(MByte)

Complexity

(GOPS/frame)

Speed

(ms/frame)

mAP bicycle scooter
icme2020_01 BUPT_MCPRL 49.20 0.90 59.50 7.35 12.89 401.21
icme2020_02 USTC-NELSLIP 44.60 0.20 56.00 6.04 11.16 141.8
icme2020_03 DD_VISION 25.60 2.00 29.40 0.86 0.52 56.18
icme2020_04 Deep Learner 49.00 0.10 62.40 45.2 56.64 1560.43
icme2020_06 nccu_vipl 38.70 0.00 48.20 187.27 285.7 287.11
icme2020_07 IBDO-AIOT 47.00 12.00 53.70 215.49 99.46 567.65
icme2020_08 ACVLab 41.10 0.10 49.90 87.46 31.82 696.72
icme2020_010 資工A 8.80 0.00 3.10 298.6 140.35 653.71
icme2020_05 Rock4Ever
icme2020_09 jummy112

Final Score

group Name Partial Score Final Score
mAP model size complexity speed
icme2020_01 BUPT_MCPRL 25 24.46 23.92 19.27 92.64
icme2020_02 USTC-NELSLIP 22.15 24.57 24.07 23.58 94.36
icme2020_03 DD_VISION 10.4 25 25 25 85.40
icme2020_04 Deep Learner 24.88 21.28 20.08 0 66.23
icme2020_06 nccu_vipl 18.5 9.35 0 21.16 49.01
icme2020_07 IBDO-AIOT 23.64 6.98 16.33 16.5 63.44
icme2020_08 ACVLab 19.99 17.73 22.26 14.35 74.33
icme2020_010 資工A 0 0 12.74 15.07 27.81
icme2020_05 Rock4Ever
icme2020_09 jummy112