We present the next generation of MobileNets based on a combination of complementary search techniques as well as a novel architecture design. MobileNetV3 is tuned to mobile phone CPUs through a combination of hardware-aware network architecture search (NAS) complemented by the NetAdapt algorithm and then subsequently improved through novel architecture advances. This paper starts the exploration of how automated search algorithms and network design can work together to harness complementary approaches improving the overall state of the art. Through this process we create two new MobileNet models for release: MobileNetV3-Large and MobileNetV3-Small which are targeted for high and low resource use cases. These models are then adapted and applied to the tasks of object detection and semantic segmentation. For the task of semantic segmentation (or any dense pixel prediction), we propose a new efficient segmentation decoder Lite Reduced Atrous Spatial Pyramid Pooling (LR-ASPP). We achieve new state of the art results for mobile classification, detection and segmentation. MobileNetV3-Large is 3.2\% more accurate on ImageNet classification while reducing latency by 15\% compared to MobileNetV2. MobileNetV3-Small is 4.6\% more accurate while reducing latency by 5\% compared to MobileNetV2. MobileNetV3-Large detection is 25\% faster at roughly the same accuracy as MobileNetV2 on COCO detection. MobileNetV3-Large LR-ASPP is 30\% faster than MobileNetV2 R-ASPP at similar accuracy for Cityscapes segmentation.

Searching for MobileNetV3.

For real time applications utilizing Deep Neural Networks (DNNs), it is critical that the models achieve high-accuracy on the target task and low-latency inference on the target computing platform. While Neural Architecture Search (NAS) has been effectively used to develop low-latency networks for image classification, there has been relatively little effort to use NAS to optimize DNN architectures for other vision tasks. In this work, we present what we believe to be the first proxyless hardware-aware search targeted for dense semantic segmentation. With this approach, we advance the state-of-the-art accuracy for latency-optimized networks on the Cityscapes semantic segmentation dataset. Our latency-optimized small SqueezeNAS network achieves 68.02% validation class mIOU with less than 35 ms inference times on the NVIDIA Xavier. Our latency-optimized large SqueezeNAS network achieves 73.62% class mIOU with less than 100 ms inference times. We demonstrate that significant performance gains are possible by utilizing NAS to find networks optimized for both the specific task and inference hardware. We also present detailed analysis comparing our networks to recent state-of-the-art architectures. The SqueezeNAS models are available for download here: https://github.com/ashaw596/squeezenas

/pdf/squeezenas-fast-neural-architecture-search-for-faster-4bn1z8hiyi.pdf

SqueezeNAS: Fast Neural Architecture Search for Faster Semantic Segmentation

High-resolution representations (HR) are essential for dense prediction tasks such as segmentation, detection, and pose estimation. Learning HR representations is typically ignored in previous Neural Architecture Search (NAS) methods that focus on image classification. This work proposes a novel NAS method, called HR-NAS, which is able to find efficient and accurate networks for different tasks, by effectively encoding multiscale contextual information while maintaining high-resolution representations. In HR-NAS, we renovate the NAS search space as well as its searching strategy. To better encode multiscale image contexts in the search space of HR-NAS, we first carefully design a lightweight transformer, whose computational complexity can be dynamically changed with respect to different objective functions and computation budgets. To maintain high-resolution representations of the learned networks, HR-NAS adopts a multi-branch architecture that provides convolutional encoding of multiple feature resolutions, inspired by HRNet [73]. Last, we proposed an efficient fine-grained search strategy to train HR-NAS, which effectively explores the search space, and finds optimal architectures given various tasks and computation resources. As shown in Fig. 1 (a), HR-NAS is capable of achieving state-of-the-art trade-offs between performance and FLOPs for three dense prediction tasks and an image classification task, given only small computational budgets. For example, HR-NAS surpasses SqueezeNAS [63] that is specially designed for semantic segmentation while improving efficiency by 45.9%. Code is available at https://github.com/dingmyu/HR-NAS.

/pdf/hr-nas-searching-efficient-high-resolution-neural-2hba7shuc5.pdf

HR-NAS: Searching Efficient High-Resolution Neural Architectures with Lightweight Transformers

D2A U-Net: Automatic segmentation of COVID-19 CT slices based on dual attention and hybrid dilated convolution.

Attention-based generative adversarial network with internal damage segmentation using thermography

One of the practical choices for making a lightweight semantic segmentation model is to combine a depth-wise separable convolution with a dilated convolution. However, the simple combination of these two methods results in an over-simplified operation which causes severe performance degradation due to loss of information contained in the feature map. To resolve this problem, we propose a new block called Concentrated-Comprehensive Convolution (C3) which applies the asymmetric convolutions before the depth-wise separable dilated convolution to compensate for the information loss due to dilated convolution. The C3 block consists of a concentration stage and a comprehensive convolution stage. The first stage uses two depth-wise asymmetric convolutions for compressed information from the neighboring pixels to alleviate the information loss. The second stage increases the receptive field by using a depth-wise separable dilated convolution from the feature map of the first stage. We applied the C3 block to various segmentation frameworks (ESPNet, DRN, ERFNet, ENet) for proving the beneficial properties of our proposed method. Experimental results show that the proposed method preserves the original accuracies on Cityscapes dataset while reducing the complexity. Furthermore, we modified ESPNet to achieve about 2% better performance while reducing the number of parameters by half and the number of FLOPs by 35% compared with the original ESPNet. Finally, experiments on ImageNet classification task show that C3 block can successfully replace dilated convolutions.

C3: Concentrated-Comprehensive Convolution and its application to semantic segmentation

The semantic segmentation requires a lot of computational cost. The dilated convolution relieves this burden of complexity by increasing the receptive field without additional parameters. For a more lightweight model, using depth-wise separable convolution is one of the practical choices. However, a simple combination of these two methods results in too sparse an operation which might cause sever performance degradation. To resolve this problem, we propose a new block of Concentrated-Comprehensive Convolution (CCC) which takes both advantages of the dilated convolution and the depth-wise separable convolution. The CCC block consists of an information concentration stage and a comprehensive convolution stage. The first stage uses two depth-wise asymmetric convolutions for compressed information from the neighboring pixels. The second stage increases the receptive field by using a depth-wise separable dilated convolution from the feature map of the first stage. By replacing the conventional ESP module with the proposed CCC module, without accuracy degradation in Cityscapes dataset, we could reduce the number of parameters by half and the number of flops by 35% compared to the ESPnet which is one of the fastest models. We further applied the CCC to other segmentation models based on dilated convolution and our method achieved comparable or higher performance with a decreased number of parameters and flops. Finally, experiments on ImageNet classification task show that CCC can successfully replace dilated convolutions.

Concentrated-Comprehensive Convolutions for lightweight semantic segmentation

The learning of the region proposal in object detection using the deep neural networks (DNN) is divided into two tasks: binary classification and bounding box regression task. However, traditional RPN (Region Proposal Network) defines these two tasks as different problems, and they are trained independently. In this paper, we propose a new region proposal learning method that considers the bounding box offset’s uncertainty in the objectness score. Our method redefines RPN to a problem of minimizing the KL-divergence, difference between the two probability distributions. We applied KLRPN, which performs region proposal using KL-Divergence, to the existing two-stage object detection framework and showed that it can improve the performance of the existing method. Experiments show that it achieves 2.6% and 2.0% AP improvements on MS COCO test-dev in Faster R-CNN with VGG-16 and R-FCN with ResNet-101 backbone, respectively.

Kl-Divergence-Based Region Proposal Network For Object Detection

In multi-object detection using neural networks, most methods train a network based on ground truth assignment, which makes the training too heuristic and complicated. In this paper, we reformulate the multi-object detection task as a problem of density estimation of bounding boxes. Instead of using a ground-truth-assignment-based method, we train a network by estimating the probability density of bounding boxes in an input image using a mixture model. For this purpose, we propose a novel network for object detection called Mixture Density Object Detector (MDOD), and the corresponding objective function for the density-estimation-based training. Unlike the ground-truth-assignment-based methods, our proposed method gets rid of the cumbersome processes of matching between ground truth boxes and their predictions as well as the heuristic anchor design. It is also free from the problem of foreground-background imbalance. We applied MDOD to MS COCO dataset. Our proposed method not only deals with multi-object detection problems in a new approach, but also improves detection performances through MDOD. Code will be available.

/pdf/density-based-object-detection-learning-bounding-boxes-5ompdsqa6p.pdf

Density-Based Object Detection: Learning Bounding Boxes without Ground Truth Assignment

In multi-object detection using neural networks, the fundamental problem is, "How should the network learn a variable number of bounding boxes in different input images?". Previous methods train a multi-object detection network through a procedure that directly assigns the ground truth bounding boxes to the specific locations of the network's output. However, this procedure makes the training of a multi-object detection network too heuristic and complicated. In this paper, we reformulate the multi-object detection task as a problem of density estimation of bounding boxes. Instead of assigning each ground truth to specific locations of network's output, we train a network by estimating the probability density of bounding boxes in an input image using a mixture model. For this purpose, we propose a novel network for object detection called Mixture Density Object Detector (MDOD), and the corresponding objective function for the density-estimation-based training. We applied MDOD to MS COCO dataset. Our proposed method not only deals with multi-object detection problems in a new approach, but also improves detection performances through MDOD. The code is available: https://github.com/yoojy31/MDOD.

Geonseok Seo

Papers

C3: Concentrated-Comprehensive Convolution and its application to semantic segmentation

Concentrated-Comprehensive Convolutions for lightweight semantic segmentation

Kl-Divergence-Based Region Proposal Network For Object Detection

Density-Based Object Detection: Learning Bounding Boxes without Ground Truth Assignment

Training Multi-Object Detector by Estimating Bounding Box Distribution for Input Image