MobileVLM V2: Faster and Stronger Baseline for Vision Language Model

We introduce MobileVLM V2, a family of significantly improved vision language models upon MobileVLM, which proves that a delicate orchestration of novel architectural design, an improved training scheme tailored for mobile VLMs, and rich high-quality dataset curation can substantially benefit VLMs' performance. Specifically, MobileVLM V2 1.7B achieves better or on-par performance on standard VLM benchmarks compared with much larger VLMs at the 3B scale. Notably, our 3B model outperforms a large variety of VLMs at the 7B+ scale. Our models will be released at https://github.com/Meituan-AutoML/MobileVLM .

MobileVLM : A Fast, Strong and Open Vision Language Assistant for Mobile Devices

We present MobileVLM, a competent multimodal vision language model (MMVLM) targeted to run on mobile devices. It is an amalgamation of a myriad of architectural designs and techniques that are mobile-oriented, which comprises a set of language models at the scale of 1.4B and 2.7B parameters, trained from scratch, a multimodal vision model that is pre-trained in the CLIP fashion, cross-modality interaction via an efficient projector. We evaluate MobileVLM on several typical VLM benchmarks. Our models demonstrate on par performance compared with a few much larger models. More importantly, we measure the inference speed on both a Qualcomm Snapdragon 888 CPU and an NVIDIA Jeston Orin GPU, and we obtain state-of-the-art performance of 21.5 tokens and 65.3 tokens per second, respectively. Our code will be made available at: https://github.com/Meituan-AutoML/MobileVLM.

Masked Autoencoders Are Robust Neural Architecture Search Learners

Neural Architecture Search (NAS) currently relies heavily on labeled data, which is both expensive and time-consuming to acquire. In this paper, we propose a novel NAS framework based on Masked Autoencoders (MAE) that eliminates the need for labeled data during the search process. By replacing the supervised learning objective with an image reconstruction task, our approach enables the robust discovery of network architectures without compromising performance and generalization ability. Additionally, we address the problem of performance collapse encountered in the widely-used Differentiable Architecture Search (DARTS) method in the unsupervised paradigm by introducing a multi-scale decoder. Through extensive experiments conducted on various search spaces and datasets, we demonstrate the effectiveness and robustness of the proposed method, providing empirical evidence of its superiority over baseline approaches.

Exploring the impact of weather on Metro demand forecasting using machine learning method

Urban rail transit provides significant comprehensive benefits such as large traffic volume and high speed, serving as one of the most important components of urban traffic construction management and congestion solution. Using real passenger flow data of an Asian subway system from April to June of 2018, this work analyzes the space-time distribution of the passenger flow using short-term traffic flow prediction. Stations are divided into four types for passenger flow forecasting, and meteorological records are collected for the same period. Then, machine learning methods with different inputs are applied and multivariate regression is performed to evaluate the improvement effect of each weather element on passenger flow forecasting of representative metro stations on hourly basis. Our results show that by inputting weather variables the precision of prediction on weekends enhanced while the performance on weekdays only improved marginally, while the contribution of different elements of weather differ. Also, different categories of stations are affected differently by weather. This study provides a possible method to further improve other prediction models, and attests to the promise of data-driven analytics for optimization of short-term scheduling in transit management.

DPUV3INT8: A Compiler View to programmable FPGA Inference Engines

We have a FPGA design, we make it fast, efficient, and tested for a few important examples. Now we must infer a general solution to deploy in the data center. Here, we describe the FPGA DPUV3INT8 design and our compiler effort. The hand-tuned SW-HW solution for Resnet50\_v1 has (close to) 2 times better images per second (throughput) than our best FPGA implementation; the compiler generalizes the hand written techniques achieving about 1.5 times better performance for the same example, the compiler generalizes the optimizations to a model zoo of networks, and it achieves 80+\% HW efficiency.

Angle-based Search Space Shrinking for Neural Architecture Search

In this work, we present a simple and general search space shrinking method, called Angle-Based search space Shrinking (ABS), for Neural Architecture Search (NAS). Our approach progressively simplifies the original search space by dropping unpromising candidates, thus can reduce difficulties for existing NAS methods to find superior architectures. In particular, we propose an angle-based metric to guide the shrinking process. We provide comprehensive evidences showing that, in weight-sharing supernet, the proposed metric is more stable and accurate than accuracy-based and magnitude-based metrics to predict the capability of child models. We also show that the angle-based metric can converge fast while training supernet, enabling us to get promising shrunk search spaces efficiently. ABS can easily apply to most of popular NAS approaches (e.g. SPOS, FariNAS, ProxylessNAS, DARTS and PDARTS). Comprehensive experiments show that ABS can dramatically enhance existing NAS approaches by providing a promising shrunk search space.

Cluster Regularized Quantization for Deep Networks Compression

Deep neural networks (DNNs) have achieved great success in a wide range of computer vision areas, but the applications to mobile devices is limited due to their high storage and computational cost. Much efforts have been devoted to compress DNNs. In this paper, we propose a simple yet effective method for deep networks compression, named Cluster Regularized Quantization (CRQ), which can reduce the presentation precision of a full-precision model to ternary values without significant accuracy drop. In particular, the proposed method aims at reducing the quantization error by introducing a cluster regularization term, which is imposed on the full-precision weights to enable them naturally concentrate around the target values. Through explicitly regularizing the weights during the re-training stage, the full-precision model can achieve the smooth transition to the low-bit one. Comprehensive experiments on benchmark datasets demonstrate the effectiveness of the proposed method.

Multi-loss-aware Channel Pruning of Deep Networks

Channel pruning, which seeks to reduce the model size by removing redundant channels, is a popular solution for deep networks compression. Existing channel pruning methods usually conduct layer-wise channel selection by directly minimizing the reconstruction error of feature maps between the baseline model and the pruned one. However, they ignore the feature and semantic distributions within feature maps and real contribution of channels to the overall performance. In this paper, we propose a new channel pruning method by explicitly using both intermediate outputs of the baseline model and the classification loss of the pruned model to supervise layer-wise channel selection. Particularly, we introduce an additional loss to encode the differences in the feature and semantic distributions within feature maps between the baseline model and the pruned one. By considering the reconstruction error, the additional loss and the classification loss at the same time, our approach can significantly improve the performance of the pruned model. Comprehensive experiments on benchmark datasets demonstrate the effectiveness of the proposed method.

Better Guider Predicts Future Better: Difference Guided Generative Adversarial Networks

Predicting the future is a fantasy but practicality work. It is the key component to intelligent agents, such as self-driving vehicles, medical monitoring devices and robotics. In this work, we consider generating unseen future frames from previous obeservations, which is notoriously hard due to the uncertainty in frame dynamics. While recent works based on generative adversarial networks (GANs) made remarkable progress, there is still an obstacle for making accurate and realistic predictions. In this paper, we propose a novel GAN based on inter-frame difference to circumvent the difficulties. More specifically, our model is a multi-stage generative network, which is named the Difference Guided Generative Adversarial Netwok (DGGAN). The DGGAN learns to explicitly enforce future-frame predictions that is guided by synthetic inter-frame difference. Given a sequence of frames, DGGAN first uses dual paths to generate meta information. One path, called Coarse Frame Generator, predicts the coarse details about future frames, and the other path, called Difference Guide Generator, generates the difference image which include complementary fine details. Then our coarse details will then be refined via guidance of difference image under the support of GANs. With this model and novel architecture, we achieve state-of-the-art performance for future video prediction on UCF-101, KITTI.

Action Machine: Rethinking Action Recognition in Trimmed Videos

Existing methods in video action recognition mostly do not distinguish human body from the environment and easily overfit the scenes and objects. In this work, we present a conceptually simple, general and high-performance framework for action recognition in trimmed videos, aiming at person-centric modeling. The method, called Action Machine, takes as inputs the videos cropped by person bounding boxes. It extends the Inflated 3D ConvNet (I3D) by adding a branch for human pose estimation and a 2D CNN for pose-based action recognition, being fast to train and test. Action Machine can benefit from the multi-task training of action recognition and pose estimation, the fusion of predictions from RGB images and poses. On NTU RGB-D, Action Machine achieves the state-of-the-art performance with top-1 accuracies of 97.2% and 94.3% on cross-view and cross-subject respectively. Action Machine also achieves competitive performance on another three smaller action recognition datasets: Northwestern UCLA Multiview Action3D, MSR Daily Activity3D and UTD-MHAD. Code will be made available.