Information
Contact Name: Carl Lemaire
Email: carl.lemaire@usherbrooke.ca
Organization: Université de Sherbrooke
Method's Name: VGG-19 (with Batch Norm)
Reference: K. Simonyan, A. Zisserman. Very Deep Convolutional Networks for Large-Scale Image Recognition. Proceedings of ICLR 2015, p.1-14.
Project Website:
Evaluation Metrics

Precision of each category :

$$ Pre_i = \frac{TP_i}{TP_i + FP_i}$$


Articulated Truck Bicycle Bus Car Motorcycle Non-motorized Vehicle Pedestrian Pickup Truck Single Unit Truck Work Van Background
0.8812 0.9015 0.9419 0.9754 0.8472 0.5993 0.8934 0.8470 0.6579 0.8452 0.9931

Recall of each category:

$$ Rec_i = \frac{TP_i}{TP_i + FN_i}$$


Articulated Truck Bicycle Bus Car Motorcycle Non-motorized Vehicle Pedestrian Pickup Truck Single Unit Truck Work Van Background
0.8856 0.7531 0.9426 0.9640 0.8626 0.3858 0.9214 0.9292 0.7047 0.7283 0.9917

Accuracy: 0.9564

$$ Acc = \frac{TP}{Number of Testing Images}$$


Mean Recall: 0.8245

$$ {mRec} = {\rm mean}({Rec_i})$$


Mean Precision: 0.8530

$$ {mPre} = {\rm mean}({Pre_i})$$


Cohen Kappa Score: 0.9323

Confusion Matrix (%):

Predicted
Articulated Truck Bicycle Bus Car Motorcycle Non-motorized Vehicle Pedestrian Pickup Truck Single Unit Truck Work Van Background
True Articulated Truck 88.56 0.04 0.58 0.89 0.04 0.58 0.08 1.43 7.07 0.27 0.46
Bicycle 0.18 75.31 0.18 2.63 5.25 0.00 14.54 0.53 0.18 0.00 1.23
Bus 0.85 0.00 94.26 0.89 0.00 0.31 0.00 0.54 1.55 1.32 0.27
Car 0.05 0.00 0.05 96.40 0.03 0.05 0.05 2.74 0.06 0.30 0.27
Motorcycle 0.00 3.64 0.00 5.05 86.26 0.20 2.63 0.81 0.40 0.20 0.81
Non-motorized Vehicle 10.96 0.00 3.88 11.19 1.14 38.58 0.68 5.02 26.48 0.91 1.14
Pedestrian 0.13 1.60 0.00 1.92 0.83 0.06 92.14 0.06 0.13 0.00 3.13
Pickup Truck 0.13 0.01 0.11 6.07 0.02 0.14 0.02 92.92 0.34 0.17 0.07
Single Unit Truck 10.00 0.08 1.95 1.80 0.23 1.48 0.08 9.61 70.47 3.98 0.31
Work Van 0.21 0.00 1.16 18.08 0.08 0.33 0.00 5.74 1.45 72.83 0.12
Background 0.13 0.00 0.05 0.47 0.01 0.03 0.09 0.03 0.02 0.02 99.17