Experimental 0.7 alpha

• Updated mxnet to 0.12.0rc1
• Include pretrained mobilenet models

Released models:

• mobilenet-608: mAP ~74.7, ~0.56s/frame on cpu
• mobilenet-512: mAP ~72.5, ~0.39s/frame on cpu
• resnet50-512: mAP ~79.1, ~1.0s/frame on cpu

Tested on i7-5557U CPU @ 3.10GHz, apple blas

Release v0.6

• Improved symbol composing, you can easily create own network based on popular classification networks.
• Optimized for multi-gpu performances.
• New models provided.

Release v0.5-beta

Release Note

Now with new iterator for faster training, together with multiple modifications to improve mAP from 72% to 78% for SSD-300 model, and 74% to 79%+ for 512 models.

Trained models:

Model	Training data	Test data	mAP
VGG16_reduced 300x300	VOC07+12 trainval	VOC07 test	77.8
VGG16_reduced 512x512	VOC07+12 trainval	VOC07 test	79.9

Models for official MXNet

For official mxnet/example/ssd(apache/mxnet@17b6aac), do not use for this repo!

1. Pretrained detection model: ssd_300_vgg16_reduced_voc0712_trainval.zip with 71.5% mAP on PASCAL VOC 2007

Release-v0.2-beta

First usable release before rewriting structures

Provided models:

1. Pretrained classification model: vgg16_reduced.zip
2. Pretrained detection model: ssd_300_voc0712.zip with 71.5% mAP on PASCAL VOC 2007