Voxel grid representation of 3D scene properties has been widely used to improve the training or rendering speed of the Neural Radiance Fields (NeRF) while at the same time achieving high synthesis quality. However, these methods accelerate the original NeRF at the expense of extra storage demand, which hinders their applications in many scenarios. To solve this limitation, we present TinyNeRF, a three-stage pipeline: frequency domain transformation, pruning and quantization that work together to reduce the storage demand of the voxel grids with little to no effects on their speed and synthesis quality. Based on the prior knowledge of visual signals sparsity in the frequency domain, we convert the original voxel grids in the frequency domain via block-wise discrete cosine transformation (DCT). Next, we apply pruning and quantization to enforce the DCT coefficients to be sparse and low-bit. Our method can be optimized from scratch in an end-to-end manner, and can typically compress the original models by 2 orders of magnitude with minimal sacrifice on speed and synthesis quality.

Direct convolution methods are now drawing increasing attention as they eliminate the additional storage demand required by indirect convolution algorithms (i.e., the transformed matrix generated by the im2col convolution algorithm). Nevertheless, the direct methods require special input-output tensor formatting, leading to extra time and memory consumption to get the desired data layout. In this article, we show that indirect convolution, if implemented properly, is able to achieve high computation performance with the help of highly optimized subroutines in matrix multiplication while avoid incurring substantial memory overhead. The proposed algorithm is called efficient convolution via blocked columnizing (ECBC). Inspired by the im2col convolution algorithm and the block algorithm of general matrix-to-matrix multiplication, we propose to conduct the convolution computation blockwisely. As a result, the tensor-to-matrix transformation process (e.g., the im2col operation) can also be done in a blockwise manner so that it only requires a small block of memory as small as the data block. Extensive experiments on various platforms and networks validate the effectiveness of ECBC, as well as the superiority of our proposed method against a set of widely used industrial-level convolution algorithms.

In this paper, we propose a novel one-hot representation for weights and activations in CNN model and demonstrate its benefits on hardware accelerator design. Specifically, rather than merely reducing the bitwidth, we quantize both weights and activations into n-bit integers that containing only one non-zero bit per value. In this way, the massive multiply and accumulates (MACs) are equivalent to additions of powers of two that can be efficiently calculated with histogram based computaitons. Experiments on the ImageNet classification task show that comparable accuracy can be obtained on our proposed One-Hot Networks (OHN) compared to conventional fixed-point networks. As case studies, we evaluate the efficacy of the one-hot data representation on two state-of-the-art CNN accelerators on FPGA, our preliminary results show that 50% and 68.5% resource saving can be achieved on DaDianNao and Laconic respectively. Besides, the one-hot optimized Laconic can further achieve an average speedup of 4.94× on AlexNet.

Face alignment is an important task in the field of multi-media. Together with the impressive progress of algorithms, various benchmark datasets have been released in recent years. Intuitively, it is meaningful to integrate multiple labeled datasets with different annotations to achieve higher performance on a target landmark detector. Although numerous efforts have been made in joint usage, there yet remain three shortages in recent works, e.g., additional computation, limitation of the markups scheme, and limited support for the regression method. To address the above problems, we proposed a novel Alternating Training Framework (ATF), which leverages similarity and diversity across multi-media sources for a more robust detector. Our framework mainly contains two sub-modules: Alternating Training with Decreasing Proportions (ATDP) and Mixed Branch Loss (mathcal LMB). In particular, ATDP trains multiple datasets simultaneously to take advantage of the diversity between them, while mathcal LMB utilizes similar landmark pairs to constrain different branches of corresponding datasets. Extensive experiments on various benchmarks show the effectiveness of our framework, and ATF is feasible for both heatmap-based network and direct coordinate regression. Specifically, the mean error even reaches 3.17 on the experiment on 300W leveraging WFLW, which significantly outperforms state-of-the-art methods. Both in an ordinary convolutional network (OCN) and HRNET, ATF achieves up to 9.96% relative improvement. Our source codes are made publicly available at https://github.com/starhiking/ATF.

Although deep learning models are highly effective for various learning tasks, their high computational costs prohibit the deployment to scenarios where either memory or computational resources are limited. In this paper, we focus on compressing and accelerating deep models with network weights represented by very small numbers of bits, referred to as extremely low bit neural network. We model this problem as a discretely constrained optimization problem. Borrowing the idea from Alternating Direction Method of Multipliers (ADMM), we decouple the continuous parameters from the discrete constraints of network, and cast the original hard problem into several subproblems. We propose to solve these subproblems using extragradient and iterative quantization algorithms that lead to considerably faster convergency compared to conventional optimization methods. Extensive experiments on image recognition and object detection verify that the proposed algorithm is more effective than state-of-the-art approaches when coming to extremely low bit neural network.

We are witnessing an explosive development and widespread application of deep neural networks (DNNs) in various fields. However, DNN models, especially a convolutional neural network (CNN), usually involve massive parameters and are computationally expensive, making them extremely dependent on high-performance hardware. This prohibits their further extensions, e.g., applications on mobile devices. In this paper, we present a quantized CNN, a unified approach to accelerate and compress convolutional networks. Guided by minimizing the approximation error of individual layer's response, both fully connected and convolutional layers are carefully quantized. The inference computation can be effectively carried out on the quantized network, with much lower memory and storage consumption. Quantitative evaluation on two publicly available benchmarks demonstrates the promising performance of our approach: with comparable classification accuracy, it achieves 4 to $6 \times $ acceleration and 15 to $20\times $ compression. With our method, accurate image classification can even be directly carried out on mobile devices within 1 s.

Convolutional neural networks (CNNs) have achieved impressive performance in a wide range of computer vision areas. However, the application on mobile devices remains intractable due to the high computation complexity. In this demo, we propose the Quantized CNN (Q-CNN), an efficient framework for CNN models, to fulfill efficient and accurate image classification on mobile devices. Our Q-CNN framework dramatically accelerates the computation and reduces the storage/memory consumption, so that mobile devices can independently run an ImageNet-scale CNN model. Experiments on the ILSVRC-12 dataset demonstrate 4~6x speed-up and 15~20x compression, with merely one percentage drop in the classification accuracy. Based on the Q-CNN framework, even mobile devices can accurately classify images within one second.

Recently, convolutional neural networks (CNN) have demonstrated impressive performance in various computer vision tasks. However, high performance hardware is typically indispensable for the application of CNN models due to the high computation complexity, which prohibits their further extensions. In this paper, we propose an efficient framework, namely Quantized CNN, to simultaneously speed-up the computation and reduce the storage and memory overhead of CNN models. Both filter kernels in convolutional layers and weighting matrices in fully-connected layers are quantized, aiming at minimizing the estimation error of each layer's response. Extensive experiments on the ILSVRC-12 benchmark demonstrate 4 ~ 6× speed-up and 15 ~ 20× compression with merely one percentage loss of classification accuracy. With our quantized CNN model, even mobile devices can accurately classify images within one second.

Recently, hashing based approximate nearest neighbor (ANN) search has attracted much attention. Extensive new algorithms have been developed and successfully applied to different applications. However, two critical problems are rarely mentioned. First, in real-world applications, the data often comes in a streaming fashion but most of existing hashing methods are batch based models. Second, when the dataset becomes huge, it is almost impossible to load all the data into memory to train hashing models. In this paper, we propose a novel approach to handle these two problems simultaneously based on the idea of data sketching. A sketch of one dataset preserves its major characters but with significantly smaller size. With a small size sketch, our method can learn hash functions in an online fashion, while needs rather low computational complexity and storage space. Extensive experiments on two large scale benchmarks and one synthetic dataset demonstrate the efficacy of the proposed method.

Recently, hashing based approximate nearest neighbors search has attracted much attention. Extensive centralized hashing algorithms have been proposed and achieved promising performance. However, due to the large scale of many applications, the data is often stored or even collected in a distributed manner. Learning hash functions by aggregating all the data into a fusion center is infeasible because of the prohibitively expensive communication and computation overhead. In this paper, we develop a novel hashing model to learn hash functions in a distributed setting. We cast a centralized hashing model as a set of subproblems with consensus constraints. We find these subproblems can be analytically solved in parallel on the distributed compute nodes. Since no training data is transmitted across the nodes in the learning process, the communication cost of our model is independent to the data size. Extensive experiments on several large scale datasets containing up to 100 million samples demonstrate the efficacy of our method.

Two tasks have been put forward in the MSR-bing Grand Challenge 2015. To address the information retrieval task, we raise and integrate a series of methods with visual features obtained by convolution neural network (CNN) models. In our experiments, we discover that the ranking strategies of Hierarchical clustering and PageRank methods are mutually complementary. Another task is fine-grained classification. In contrast to basic-level recognition, fine-grained classification aims to distinguish between different breeds or species or product models, and often requires distinctions that must be conditioned on the object pose for reliable identification. Current state-of-the-art techniques rely heavily upon the use of part annotations, while the bing datasets suffer both abundance of part annotations and dirty background. In this paper, we propose a CNN-based feature representation for visual recognition only using image-level information. Our CNN model is pre-trained on a collection of clean datasets and fine-tuned on the bing datasets. Furthermore, a multi-scale training strategy is adopted by simply resizing the input images into different scales and then merging the soft-max posteriors. We then implement our method into a unified visual recognition system on Microsoft cloud service. Finally, our solution achieved top performance in both tasks of the contest

Due to the ability to preserve semantic similarity in Hamming space, supervised hashing has been extensively studied recently. Most existing approaches encourage two dissimilar samples to have maximum Hamming distance. This may lead to an unexpected consequence that two unnecessarily similar samples would have the same code if they are both dissimilar with another sample. Besides, in existing methods, all labeled pairs are treated with equal importance without considering the semantic gap, which is not conducive to thoroughly leverage the supervised information. We present a general framework for supervised hashing to address the above two limitations. We do not toughly require a dissimilar pair to have maximum Hamming distance. Instead, a soft constraint which can be viewed as a regularization to avoid over-fitting is utilized. Moreover, we impose different weights to different training pairs, and these weights can be automatically adjusted in the learning process. Experiments on two benchmarks show that the proposed method can easily outperform other state-of-the-art methods.

Image matching is one of the most challenging stages in 3D reconstruction, which usually occupies half of computational cost and inaccurate matching may lead to failure of reconstruction. Therefore, fast and accurate image matching is very crucial for 3D reconstruction. In this paper, we proposed a Cascade Hashing strategy to speed up the image matching. In order to accelerate the image matching, the proposed Cascade Hashing method is designed to be three-layer structure: hashing lookup, hashing remapping, and hashing ranking. Each layer adopts different measures and filtering strategies, which is demonstrated to be less sensitive to noise. Extensive experiments show that image matching can be accelerated by our approach in hundreds times than brute force matching, even achieves ten times or more than Kd-tree based matching while retaining comparable accuracy.

Due to the fast query speed and low storage cost, hashing based approximate nearest neighbor search methods have attracted much attention recently. Many state of the art methods are based on eigenvalue decomposition. In these approaches, the information caught in different dimensions is unbalanced and generally most of the information is contained in the top eigenvectors. We demonstrate that this leads to an unexpected phenomenon that longer hashing code does not necessarily yield better performance. In this work, we introduce a random subspace strategy to address this limitation. At first, a small fraction of the whole feature space is randomly sampled to train the hashing algorithms each time and only the top eigenvectors are kept to generate one piece of short code. This process will be repeated several times and then the obtained many pieces of short codes are concatenated into one piece of long code. Theoretical analysis and experiments on two benchmarks confirm the effectiveness of the proposed strategy for hashing.

Collaborative filtering with implicit feedbacks has been steadily receiving more attention, since the abundant implicit feedbacks are more easily collected while explicit feedbacks are not necessarily always available. Several recent work address this problem well utilizing pairwise ranking method with a fundamental assumption that a user prefers items with positive feedbacks to the items without observed feedbacks, which also implies that the items without observed feedbacks are treated equally without distinction. However, users have their own preference on different items with different degrees which can be modeled into a ranking relationship. In this paper, we exploit this prior information of a user's preference from the nearest neighbor set by the neighbors' implicit feedbacks, which can split items into different item groups with specific ranking relations. We propose a novel PRIGP(Personalized Ranking with Item Group based Pairwise preference learning) algorithm to integrate item based pairwise preference and item group based pairwise preference into the same framework. Experimental results on three real-world datasets demonstrate the proposed method outperforms the competitive baselines on several ranking-oriented evaluation metrics.