OpenGM is a C++ template library for defining discrete graphical models and performing inference on these models, using a wide range of state-of-the-art algorithms. No restrictions are imposed on the factor graph to allow for higher-order factors and arbitrary neighborhood structures. Large models with repetitive structure are handled efficiently because (i) functions that occur repeatedly need to be stored only once, and (ii) distinct functions can be implemented differently, using different encodings alongside each other in the same model. Several parametric functions (e.g. metrics), sparse and dense value tables are provided and so is an interface for custom C++ code. Algorithms are separated by design from the representation of graphical models and are easily exchangeable. OpenGM, its algorithms, HDF5 file format and command line tools are modular and extendible.
The ability to build maps is a key functionality for the majority of mobile robots. A central ingredient to most mapping systems is the registration or alignment of the recorded sensor data. In this paper, we present a general methodology for photometric registration that can deal with multiple different cues. We provide examples for registering RGBD as well as 3D LIDAR data. In contrast to popular point cloud registration approaches such as ICP our method does not rely on explicit data association and exploits multiple modalities such as raw range and image data streams. Color, depth, and normal information are handled in an uniform manner and the registration is obtained by minimizing the pixel-wise difference between two multi-channel images. We developed a flexible and general framework and implemented our approach inside that framework. We also released our implementation as open source C++ code. The experiments show that our approach allows for an accurate registration of the sensor data without requiring an explicit data association or model-specific adaptations to datasets or sensors. Our approach exploits the different cues in a natural and consistent way and the registration can be done at framerate for a typical range or imaging sensor.
Towards the vision of translating code that implements an algorithm from one programming language into another, this paper proposes an approach for automated program classification using bilateral tree-based convolutional neural networks (BiTBCNNs). It is layered on top of two tree-based convolutional neural networks (TBCNNs), each of which recognizes the algorithm of code written in an individual programming language. The combination layer of the networks recognizes the similarities and differences among code in different programming languages. The BiTBCNNs are trained using the source code in different languages but known to implement the same algorithms and/or functionalities. For a preliminary evaluation, we use 3591 Java and 3534 C++ code snippets from 6 algorithms we crawled systematically from GitHub. We obtained over 90% accuracy in the cross-language binary classification task to tell whether any given two code snippets implement the same algorithm. Also, for the algorithm classification task, i.e., to predict which one of the six algorithm labels is implemented by an arbitrary C++ code snippet, we achieved over 80% precision.
Traditional approaches to stereo visual SLAM rely on point features to estimate the camera trajectory and build a map of the environment. In low-textured environments, though, it is often difficult to find a sufficient number of reliable point features and, as a consequence, the performance of such algorithms degrades. This paper proposes PL-SLAM, a stereo visual SLAM system that combines both points and line segments to work robustly in a wider variety of scenarios, particularly in those where point features are scarce or not well-distributed in the image. PL-SLAM leverages both points and segments at all the instances of the process: visual odometry, keyframe selection, bundle adjustment, etc. We contribute also with a loop closure procedure through a novel bag-of-words approach that exploits the combined descriptive power of the two kinds of features. Additionally, the resulting map is richer and more diverse in 3D elements, which can be exploited to infer valuable, high-level scene structures like planes, empty spaces, ground plane, etc. (not addressed in this work). Our proposal has been tested with several popular datasets (such as KITTI and EuRoC), and is compared to state of the art methods like ORB-SLAM, revealing a more robust performance in most of the experiments, while still running in real-time. An open source version of the PL-SLAM C++ code will be released for the benefit of the community.
This article is about an implementation and compilation technique that is used in RAW-Feldspar which is a complete rewrite of the Feldspar embedded domain-specific language (EDSL) (Axelsson et al. 2010). Feldspar is high-level functional language that generates efficient C code to run on embedded targets. The gist of the technique presented in this post is the following: rather writing a back end that converts pure Feldspar expressions directly to C, we translate them to a low-level monadic EDSL. From the low-level EDSL, C code is then generated. This approach has several advantages: 1. The translation is simpler to write than a complete C back end. 2. The translation is between two typed EDSLs, which rules out many potential errors. 3. The low-level EDSL is reusable and can be shared between several high-level EDSLs. Although the article contains a lot of code, most of it is in fact reusable. As mentioned in Discussion, we can write the same implementation in less than 50 lines of code using generic libraries that we have developed to support Feldspar.
대규모 언어 모델(LLM)은 일상적인 코딩 작업에서 강한 성능을 낼 수 있지만, 프로그램 의미에 대한 비자명한(non-trivial) 추론을 요구하는 복잡한 작업에서는 실패할 수 있다. LLM이 이러한 작업을 풀도록 가르치기 위한 학습 예시를 마련하는 것은 어려울 수 있다. 본 연구에서는 의미적 비동치 게임(semantic inequivalence game, SInQ)에 기반해 코드 추론 학습 데이터를 합성적으로 생성하는 방법을 탐구한다. 생성자 에이전트는 실제 프로그래밍 과제 데이터셋에서 파생된, 의미적으로 구별되는 프로그램 변형을 만들고, 평가자 에이전트는 두 변형이 서로 다르게 동작하는 입력 예시를 찾아야 한다. 에이전트들은 준적대적으로(semi-adversarially) 서로를 학습시키며 코드의 기저 논리를 이해하는 능력을 향상시킨다. 우리는 코드 생성 및 이해 벤치마크 여러 곳에서 접근법을 평가했으며, 여기에는 교차 언어 취약점 탐지(Lu et al., 2021)가 포함된다. 이 설정에서 우리 방법은 오직 Python 코드로만 학습했음에도 C/C++ 코드의 취약점 탐지를 개선하며, 까다로운 Python 내장 식별자 교체 벤치마크(Miceli Barone et al., 2023)에서는 최신 LLM이 여전히 이 벤치마크에서 어려움을 겪는 반면 우리 접근법이 상당한 개선을 보여준다. 실험 재현에 필요한 코드와 LLM 파인튜닝에 사용할 수 있는 생성된 합성 데이터를 공개한다.