1# Hosted models 2 3The following is an incomplete list of pre-trained models optimized to work with 4TensorFlow Lite. 5 6To get started choosing a model, visit <a href="../models">Models</a>. 7 8Note: The best model for a given application depends on your requirements. For 9example, some applications might benefit from higher accuracy, while others 10require a small model size. You should test your application with a variety of 11models to find the optimal balance between size, performance, and accuracy. 12 13## Image classification 14 15For more information about image classification, see 16<a href="../image_classification/overview.md">Image classification</a>. 17 18### Quantized models 19 20<a href="../performance/post_training_quantization.md">Quantized</a> image 21classification models offer the smallest model size and fastest performance, at 22the expense of accuracy. 23 24Model name | Paper and model | Model size | Top-1 accuracy | Top-5 accuracy | TF Lite performance 25--------------------------- | :-------------------------------------------------------------------------------------------------------------------------------------------------------: | ---------: | -------------: | -------------: | ------------------: 26Mobilenet_V1_0.25_128_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.25_128_quant.tgz) | 0.5 Mb | 39.5% | 64.4% | 3.7 ms 27Mobilenet_V1_0.25_160_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.25_160_quant.tgz) | 0.5 Mb | 42.8% | 68.1% | 5.5 ms 28Mobilenet_V1_0.25_192_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.25_192_quant.tgz) | 0.5 Mb | 45.7% | 70.8% | 7.9 ms 29Mobilenet_V1_0.25_224_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.25_224_quant.tgz) | 0.5 Mb | 48.2% | 72.8% | 10.4 ms 30Mobilenet_V1_0.50_128_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.5_128_quant.tgz) | 1.4 Mb | 54.9% | 78.1% | 8.8 ms 31Mobilenet_V1_0.50_160_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.5_160_quant.tgz) | 1.4 Mb | 57.2% | 80.5% | 13.0 ms 32Mobilenet_V1_0.50_192_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.5_192_quant.tgz) | 1.4 Mb | 59.9% | 82.1% | 18.3 ms 33Mobilenet_V1_0.50_224_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.5_224_quant.tgz) | 1.4 Mb | 61.2% | 83.2% | 24.7 ms 34Mobilenet_V1_0.75_128_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.75_128_quant.tgz) | 2.6 Mb | 55.9% | 79.1% | 16.2 ms 35Mobilenet_V1_0.75_160_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.75_160_quant.tgz) | 2.6 Mb | 62.4% | 83.7% | 24.3 ms 36Mobilenet_V1_0.75_192_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.75_192_quant.tgz) | 2.6 Mb | 66.1% | 86.2% | 33.8 ms 37Mobilenet_V1_0.75_224_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_0.75_224_quant.tgz) | 2.6 Mb | 66.9% | 86.9% | 45.4 ms 38Mobilenet_V1_1.0_128_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_128_quant.tgz) | 4.3 Mb | 63.3% | 84.1% | 24.9 ms 39Mobilenet_V1_1.0_160_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_160_quant.tgz) | 4.3 Mb | 66.9% | 86.7% | 37.4 ms 40Mobilenet_V1_1.0_192_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_192_quant.tgz) | 4.3 Mb | 69.1% | 88.1% | 51.9 ms 41Mobilenet_V1_1.0_224_quant | [paper](https://arxiv.org/pdf/1712.05877.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_08_02/mobilenet_v1_1.0_224_quant.tgz) | 4.3 Mb | 70.0% | 89.0% | 70.2 ms 42Mobilenet_V2_1.0_224_quant | [paper](https://arxiv.org/abs/1806.08342), [tflite&pb](http://download.tensorflow.org/models/tflite_11_05_08/mobilenet_v2_1.0_224_quant.tgz) | 3.4 Mb | 70.8% | 89.9% | 80.3 ms 43Inception_V1_quant | [paper](https://arxiv.org/abs/1409.4842), [tflite&pb](http://download.tensorflow.org/models/inception_v1_224_quant_20181026.tgz) | 6.4 Mb | 70.1% | 89.8% | 154.5 ms 44Inception_V2_quant | [paper](https://arxiv.org/abs/1512.00567), [tflite&pb](http://download.tensorflow.org/models/inception_v2_224_quant_20181026.tgz) | 11 Mb | 73.5% | 91.4% | 235.0 ms 45Inception_V3_quant | [paper](https://arxiv.org/abs/1806.08342),[tflite&pb](http://download.tensorflow.org/models/tflite_11_05_08/inception_v3_quant.tgz) | 23 Mb | 77.5% | 93.7% | 637 ms 46Inception_V4_quant | [paper](https://arxiv.org/abs/1602.07261), [tflite&pb](http://download.tensorflow.org/models/inception_v4_299_quant_20181026.tgz) | 41 Mb | 79.5% | 93.9% | 1250.8 ms 47 48Note: The model files include both TF Lite FlatBuffer and Tensorflow frozen 49Graph. 50 51Note: Performance numbers were benchmarked on Pixel-2 using single thread large 52core. Accuracy numbers were computed using the 53[TFLite accuracy tool](https://github.com/tensorflow/tensorflow/tree/master/tensorflow/lite/tools/accuracy/ilsvrc). 54 55### Floating point models 56 57Floating point models offer the best accuracy, at the expense of model size and 58performance. <a href="../performance/gpu.md">GPU acceleration</a> requires the 59use of floating point models. 60 61Model name | Paper and model | Model size | Top-1 accuracy | Top-5 accuracy | TF Lite performance | Tensorflow performance 62--------------------- | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------: | ---------: | -------------: | -------------: | ------------------: | ---------------------: 63DenseNet | [paper](https://arxiv.org/abs/1608.06993), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite/model_zoo/upload_20180427/densenet_2018_04_27.tgz) | 43.6 Mb | 64.2% | 85.6% | 894 ms | 1262 ms 64SqueezeNet | [paper](https://arxiv.org/abs/1602.07360), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite/model_zoo/upload_20180427/squeezenet_2018_04_27.tgz) | 5.0 Mb | 49.0% | 72.9% | 224 ms | 255 ms 65NASNet mobile | [paper](https://arxiv.org/abs/1707.07012), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite/model_zoo/upload_20180427/nasnet_mobile_2018_04_27.tgz) | 21.4 Mb | 73.9% | 91.5% | 261 ms | 389 ms 66NASNet large | [paper](https://arxiv.org/abs/1707.07012), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite/model_zoo/upload_20180427/nasnet_large_2018_04_27.tgz) | 355.3 Mb | 82.6% | 96.1% | 6697 ms | 7940 ms 67ResNet_V2_101 | [paper](https://arxiv.org/abs/1603.05027), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite_11_05_08/resnet_v2_101.tgz) | 178.3 Mb | 76.8% | 93.6% | 1880 ms | 1970 ms 68Inception_V3 | [paper](http://arxiv.org/abs/1512.00567), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite/model_zoo/upload_20180427/inception_v3_2018_04_27.tgz) | 95.3 Mb | 77.9% | 93.8% | 1433 ms | 1522 ms 69Inception_V4 | [paper](http://arxiv.org/abs/1602.07261), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite/model_zoo/upload_20180427/inception_v4_2018_04_27.tgz) | 170.7 Mb | 80.1% | 95.1% | 2986 ms | 3139 ms 70Inception_ResNet_V2 | [paper](https://arxiv.org/abs/1602.07261), [tflite&pb](https://storage.googleapis.com/download.tensorflow.org/models/tflite/model_zoo/upload_20180427/inception_resnet_v2_2018_04_27.tgz) | 121.0 Mb | 77.5% | 94.0% | 2731 ms | 2926 ms 71Mobilenet_V1_0.25_128 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.25_128.tgz) | 1.9 Mb | 41.4% | 66.2% | 6.2 ms | 13.0 ms 72Mobilenet_V1_0.25_160 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.25_160.tgz) | 1.9 Mb | 45.4% | 70.2% | 8.6 ms | 19.5 ms 73Mobilenet_V1_0.25_192 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.25_192.tgz) | 1.9 Mb | 47.1% | 72.0% | 12.1 ms | 27.8 ms 74Mobilenet_V1_0.25_224 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.25_224.tgz) | 1.9 Mb | 49.7% | 74.1% | 16.2 ms | 37.3 ms 75Mobilenet_V1_0.50_128 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.5_128.tgz) | 5.3 Mb | 56.2% | 79.3% | 18.1 ms | 29.9 ms 76Mobilenet_V1_0.50_160 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.5_160.tgz) | 5.3 Mb | 59.0% | 81.8% | 26.8 ms | 45.9 ms 77Mobilenet_V1_0.50_192 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.5_192.tgz) | 5.3 Mb | 61.7% | 83.5% | 35.6 ms | 65.3 ms 78Mobilenet_V1_0.50_224 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.5_224.tgz) | 5.3 Mb | 63.2% | 84.9% | 47.6 ms | 164.2 ms 79Mobilenet_V1_0.75_128 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.75_128.tgz) | 10.3 Mb | 62.0% | 83.8% | 34.6 ms | 48.7 ms 80Mobilenet_V1_0.75_160 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.75_160.tgz) | 10.3 Mb | 65.2% | 85.9% | 51.3 ms | 75.2 ms 81Mobilenet_V1_0.75_192 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.75_192.tgz) | 10.3 Mb | 67.1% | 87.2% | 71.7 ms | 107.0 ms 82Mobilenet_V1_0.75_224 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_0.75_224.tgz) | 10.3 Mb | 68.3% | 88.1% | 95.7 ms | 143.4 ms 83Mobilenet_V1_1.0_128 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_128.tgz) | 16.9 Mb | 65.2% | 85.7% | 57.4 ms | 76.8 ms 84Mobilenet_V1_1.0_160 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_160.tgz) | 16.9 Mb | 68.0% | 87.7% | 86.0 ms | 117.7 ms 85Mobilenet_V1_1.0_192 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_192.tgz) | 16.9 Mb | 69.9% | 89.1% | 118.6 ms | 167.3 ms 86Mobilenet_V1_1.0_224 | [paper](https://arxiv.org/pdf/1704.04861.pdf), [tflite&pb](http://download.tensorflow.org/models/mobilenet_v1_2018_02_22/mobilenet_v1_1.0_224.tgz) | 16.9 Mb | 71.0% | 89.9% | 160.1 ms | 224.3 ms 87Mobilenet_V2_1.0_224 | [paper](https://arxiv.org/pdf/1801.04381.pdf), [tflite&pb](http://download.tensorflow.org/models/tflite_11_05_08/mobilenet_v2_1.0_224.tgz) | 14.0 Mb | 71.8% | 90.6% | 117 ms | 88 89### AutoML mobile models 90 91The following image classification models were created using 92<a href="https://cloud.google.com/automl/">Cloud AutoML</a>. 93 94Model Name | Paper and model | Model size | Top-1 accuracy | Top-5 accuracy | TF Lite performance 95---------------- | :------------------------------------------------------------------------------------------------------------------------------------------------------------: | ---------: | -------------: | -------------: | ------------------: 96MnasNet_0.50_224 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_0.5_224_09_07_2018.tgz) | 8.5 Mb | 68.03% | 87.79% | 37 ms 97MnasNet_0.75_224 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_0.75_224_09_07_2018.tgz) | 12 Mb | 71.72% | 90.17% | 61 ms 98MnasNet_1.0_96 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_1.0_96_09_07_2018.tgz) | 17 Mb | 62.33% | 83.98% | 23 ms 99MnasNet_1.0_128 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_1.0_128_09_07_2018.tgz) | 17 Mb | 67.32% | 87.70% | 34 ms 100MnasNet_1.0_160 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_1.0_160_09_07_2018.tgz) | 17 Mb | 70.63% | 89.58% | 51 ms 101MnasNet_1.0_192 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_1.0_192_09_07_2018.tgz) | 17 Mb | 72.56% | 90.76% | 70 ms 102MnasNet_1.0_224 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_1.0_224_09_07_2018.tgz) | 17 Mb | 74.08% | 91.75% | 93 ms 103MnasNet_1.3_224 | [paper](https://arxiv.org/abs/1807.11626), [tflite&pb](https://storage.cloud.google.com/download.tensorflow.org/models/tflite/mnasnet_1.3_224_09_07_2018.tgz) | 24 Mb | 75.24% | 92.55% | 152 ms 104 105Note: Performance numbers were benchmarked on Pixel-1 using single thread large 106BIG core. 107 108## Object detection 109 110For more information about object detection, see 111<a href="../models/object_detection/overview.md">Object detection</a>. 112 113The object detection model we currently host is 114**coco_ssd_mobilenet_v1_1.0_quant_2018_06_29**. 115 116<a class="button button-primary" href="http://storage.googleapis.com/download.tensorflow.org/models/tflite/coco_ssd_mobilenet_v1_1.0_quant_2018_06_29.zip">Download 117model and labels</a> 118 119## Pose estimation 120 121For more information about pose estimation, see 122<a href="../models/pose_estimation/overview.md">Pose estimation</a>. 123 124The pose estimation model we currently host is 125**multi_person_mobilenet_v1_075_float**. 126 127<a class="button button-primary" href="https://storage.googleapis.com/download.tensorflow.org/models/tflite/gpu/multi_person_mobilenet_v1_075_float.tflite">Download 128model</a> 129 130## Image segmentation 131 132For more information about image segmentation, see 133<a href="../models/segmentation/overview.md">Segmentation</a>. 134 135The image segmentation model we currently host is **deeplabv3_257_mv_gpu**. 136 137<a class="button button-primary" href="https://storage.googleapis.com/download.tensorflow.org/models/tflite/gpu/deeplabv3_257_mv_gpu.tflite">Download 138model</a> 139 140## Smart reply 141 142For more information about smart reply, see 143<a href="../models/smart_reply/overview.md">Smart reply</a>. 144 145The smart reply model we currently host is **smartreply_1.0_2017_11_01**. 146 147<a class="button button-primary" href="https://storage.googleapis.com/download.tensorflow.org/models/smartreply_1.0_2017_11_01.zip">Download 148model</a> 149