The networking integrity of the Mobilenet system uses default mechanisms such as a batchnorm and ReLU nonlinearity operation (excluding the final fully connected layer possessing no non-linearity and utilizes a softmax layer for classification) for dataset processing. The compact nature of the SSD Mobilenet framework is associated with a reduction in both parameters and Mult-adds. Although the neural network structure of SSD Mobilenet is notorious for minimal accuracy in relation to the Faster R-CNN Inception models, an intensively trained Mobilenet model can equate accuracy and precision on output object detection. Nevertheless, preliminary average precision scores (mAP) on small background objects reveal reduced values in relation to the inception models. Table 1.0 exhibits the pretrained COCO-Tensorflow models and their relationships in speed (ms) and mAP(^1):
Coco Model Name | Speed (ms) | mAP^[1] |
SSD Mobilenet V1 | 30 | 21 |
SSD Mobilenet V2 | 31 | 22 |
SSD Lite mobilenet v2 | 27 | 22 |
SSD inception v2 | 42 | 24 |
Faster RCNN Inception v2 | 58 | 28 |
Faster RCNN ResNet 50 | 89 | 30 |
Faster RCNN ResNet 50 Low Proposals | 64 | 30 |
RFCN-ResNet 101 | 92 | 30 |
Faster RCNN ResNet 101 | 106 | 32 |
Faster RCNN ResNet 101 Low Proposals | 82 | 32 |
Faster RCNN Inception ResNet Atrous V2 | 620 | 37 |
Faster RCNN Inception ResNet Atrous v2 Low Proposals | 241 | 37 |
+ | 1833 | 43 |
Faster RCNN Nas Low Proposals | 540 | 43 |
Table 1.0 (above) and Figure 1.4 (below): Relationships between typical object detection API models and their associated speeds and measures of average precision. The models of interest, SSD Mobilenet and Faster RCNN Inception, exhibit the most compatibility for future deployment with a coupling of efficiency and precision.