Arbitrary style transfer models in TF Lite with InceptionV3 backbone

This release contains TF Lite models that are based on an InceptionV3 backbone producing higher quality images. The higher quality comes at the expense of increased latency, though. A brief overview of the structure of the models is available here.

The checkpoints were obtained using the code that comes from Magenta's arbitrary image stylization work.

Note: These TF Lite models are populated with required metadata that would make it super easy to import them in Android Studio. Know more about metadata generation for TF Lite models from here.

float16 and int8 quantized TFLite models of the MobileNetV2 variants of DeepLab

Contains float16 and int8 quantized TFLite models converted from mobilenetv2_dm05_coco_voc_trainval and mobilenetv2_coco_voctrainval checkpoints gathered from https://github.com/tensorflow/models/blob/master/research/deeplab/g3doc/model_zoo.md. The original model was trained on the PASCAL VOC 2012 dataset.

TUNIT model files in .onnx, .pb, and .tflite

Contains the onnx model files frozen inference graphs, TF Lite model files of the TUNIT model. Main GitHub repository: https://github.com/clovaai/tunit. The original model files are in PyTorch.

animalFaces10_1_00.zip has been taken from here.

DeepLabV3 TFLite model files

Contains TFLite models converted from the checkpoints gathered from https://github.com/tensorflow/models/blob/master/research/deeplab/g3doc/model_zoo.md. Models cover the following datasets:

• PASCAL VOC 2012
• CityScapes
• ADE20k

All of these models are also available on TensorFlow Hub: https://tfhub.dev/s?publisher=sayakpaul.