The googlenet-v4 model is the most recent of the Inception family of models designed to perform image classification. Like the other Inception models, the googlenet-v4 model has been pretrained on the ImageNet image database. For details about this family of models, check out the paper.
The model input is a blob that consists of a single image of 1x3x299x299 in BGR order. The BGR mean values need to be subtracted as follows: [128.0,128.0,128.0] before passing the image blob into the network. In addition, values must be divided by 0.0078125.
The model output for googlenet-v4 is the typical object classifier output for the 1000 different classifications matching those in the ImageNet database.
| Metric | Value |
|---|---|
| Type | Classification |
| GFLOPs | 24.584 |
| MParams | 42.648 |
| Source framework | Caffe* |
See https://github.com/soeaver/caffe-model.
Image, name - data, shape - 1,3,299,299, format is B,C,H,W where:
B- batch sizeC- channelH- heightW- width
Channel order is BGR.
Mean values - [128.0,128.0,128.0], scale value - 128.0
Image, name - data, shape - 1,3,299,299, format is B,C,H,W where:
B- batch sizeC- channelH- heightW- width
Channel order is BGR
Object classifier according to ImageNet classes, name - prob, shape - 1,1000, output data format is B,C where:
B- batch sizeC- Predicted probabilities for each class in [0, 1] range
Object classifier according to ImageNet classes, name - prob, shape - 1,1000, output data format is B,C where:
B- batch sizeC- Predicted probabilities for each class in [0, 1] range
https://raw.githubusercontent.com/soeaver/caffe-model/master/LICENSE