*Total -- 6,697.83kb -> 5,626.72kb (15.99%)

/PyTorch/Detection/SSD/img/ssd_diagram.png -- 51.71kb -> 26.76kb (48.25%)
/PyTorch/Translation/Transformer/BLEU.png -- 50.02kb -> 32.07kb (35.89%)
/TensorFlow/LanguageModeling/BERT/data/images/bert_pipeline.png -- 207.54kb -> 133.29kb (35.77%)
/TensorFlow/Segmentation/UNet_Industrial/images/unet.png -- 668.53kb -> 447.82kb (33.01%)
/PyTorch/Classification/RN50v1.5/img/DGX2_250_loss.png -- 30.17kb -> 21.27kb (29.5%)
/PyTorch/Translation/GNMT/img/training_accuracy.png -- 114.66kb -> 81.28kb (29.11%)
/PyTorch/Translation/GNMT/img/diagram.png -- 350.71kb -> 251.50kb (28.29%)
/TensorFlow/Translation/GNMT/img/diagram.png -- 350.71kb -> 251.50kb (28.29%)
/PyTorch/Classification/RN50v1.5/img/DGX2_250_top1.png -- 36.13kb -> 26.65kb (26.22%)
/PyTorch/SpeechSynthesis/Tacotron2/img/waveglow_arch.png -- 88.12kb -> 65.50kb (25.68%)
/TensorFlow/Segmentation/UNet_Medical/images/unet.png -- 100.85kb -> 75.65kb (24.98%)
/PyTorch/Classification/RN50v1.5/img/DGX2_250_top5.png -- 33.82kb -> 25.56kb (24.44%)
/PyTorch/SpeechSynthesis/Tacotron2/img/tacotron2_fp16_loss.png -- 28.16kb -> 21.59kb (23.34%)
/PyTorch/Segmentation/MaskRCNN/img/accuracy_curve.png -- 54.72kb -> 42.27kb (22.75%)
/PyTorch/SpeechSynthesis/Tacotron2/img/waveglow_fp32_loss.png -- 50.44kb -> 39.14kb (22.39%)
/PyTorch/Translation/Transformer/transformer.png -- 191.70kb -> 154.77kb (19.27%)
/PyTorch/SpeechRecognition/Jasper/images/jasper_dense_residual.png -- 246.20kb -> 201.10kb (18.32%)
/TensorFlow/Recommendation/NCF/img/hr_histogram.png -- 27.49kb -> 22.52kb (18.1%)
/PyTorch/Recommendation/NCF/img/ncf_diagram.png -- 174.41kb -> 144.95kb (16.89%)
/PyTorch/SpeechSynthesis/Tacotron2/img/waveglow_fp16_loss.png -- 69.51kb -> 58.21kb (16.27%)
/PyTorch/SpeechRecognition/Jasper/images/jasper_model.png -- 295.17kb -> 247.51kb (16.15%)
/TensorFlow/Recommendation/NCF/img/ncf_diagram.png -- 128.24kb -> 107.53kb (16.15%)
/PyTorch/Translation/GNMT/img/training_loss.png -- 295.10kb -> 251.62kb (14.74%)
/PyTorch/Segmentation/MaskRCNN/img/loss_curve.png -- 40.23kb -> 34.63kb (13.93%)
/PyTorch/SpeechSynthesis/Tacotron2/img/tacotron2_arch.png -- 196.83kb -> 171.84kb (12.7%)
/TensorFlow/Segmentation/UNet_Industrial/images/dagm2007.png -- 596.90kb -> 521.94kb (12.56%)
/TensorFlow/Recommendation/NCF/img/dgx1v16_curve.png -- 41.65kb -> 36.54kb (12.27%)
/PyTorch/Recommendation/NCF/img/dgx1v16_curve.png -- 41.65kb -> 36.54kb (12.27%)
/TensorFlow/Recommendation/NCF/img/dgx1v_32_curve.png -- 41.58kb -> 36.57kb (12.05%)
/TensorFlow/Recommendation/NCF/img/dgx2_curve.png -- 39.15kb -> 34.70kb (11.37%)
/PyTorch/Recommendation/NCF/img/dgx2_curve.png -- 39.15kb -> 34.70kb (11.37%)
/PyTorch/SpeechSynthesis/Tacotron2/img/tacotron2_fp32_loss.png -- 29.15kb -> 26.11kb (10.45%)
/PyTorch/SpeechSynthesis/Tacotron2/img/tacotron2_amp_loss.png -- 30.52kb -> 27.33kb (10.44%)
/PyTorch/Recommendation/NCF/img/hr_histogram.png -- 29.36kb -> 27.51kb (6.29%)
/PyTorch/Recommendation/NCF/img/dgx1v_32_curve.png -- 40.93kb -> 38.79kb (5.23%)
/PyTorch/LanguageModeling/BERT/images/model.png -- 56.50kb -> 55.15kb (2.39%)
/PyTorch/Segmentation/MaskRCNN/pytorch/demo/demo_e2e_mask_rcnn_X_101_32x8d_FPN_1x.png -- 894.07kb -> 886.23kb (0.88%)
/PyTorch/Segmentation/MaskRCNN/pytorch/demo/demo_e2e_mask_rcnn_R_50_FPN_1x.png -- 898.65kb -> 890.92kb (0.86%)
/TensorFlow/LanguageModeling/BERT/data/images/trtis_bs_8.png -- 16.95kb -> 16.83kb (0.7%)
/TensorFlow/LanguageModeling/BERT/data/images/trtis_dynamic.png -- 20.43kb -> 20.34kb (0.42%)

Signed-off-by: ImgBotApp <ImgBotHelp@gmail.com>