The outcomes of the third Challenge on HArvesting Raw Tables from Infographics (ICPR 2022 CHART-Infographics) are presented in this work. Recognizing charts is a difficult process which we divided into the following task: Chart Image Classification (Task 1), Text Detection and Recognition (Task 2), Text Role Classification (Task 3), Axis Analysis (Task 4), Legend Analysis (Task 5), Plot Element Detection and Classification (Task 6.a), Data Extraction (Task 6.b), and End-to-End Data Extraction (Task 7). We have provided a novel dataset for training reusing all available data from previous challenges, and we also provide a brand new testing dataset for the evaluation of submissions. Both datasets were constructed by manually annotating charts extracted from the Open Access section of the PubMed Central. A total of 9 teams registered out of which 5 submitted results for different tasks of the challenge. Many submissions are based on state-of-the-art methods from computer vision, but the final scores imply that more work will be required to solve the chart recognition problem. The data, annotation tools, and evaluation scripts have been publicly released for academic use.

Representational Learning in the form of high dimensional embeddings have been used for multiple pattern recognition applications. There has been a significant interest in building embedding based systems for learning representations in the mathematical domain. At the same time, retrieval of structured information such as mathematical expressions is an important need for modern IR systems. In this work, our motivation is to introduce a robust framework for learning representations for similarity based retrieval of mathematical expressions. Given a query by example, the embedding can find the closest matching expression as a function of euclidean distance between them. We leverage recent advancements in image-based and graph-based deep learning algorithms to learn our similarity embeddings. We do this first, by using unimodal encoders in graph space and image space and then, a multi-modal combination of the same. To overcome the lack of training data, we force the networks to learn a deep metric using triplets generated with a heuristic scoring function. We also adopt a custom strategy for mining hard samples to train our neural networks. Our system produces rankings similar to those generated by the original scoring function, but using only a fraction of the time. Our results establish the viability of using such a multi-modal embedding for this task.

Minimum mean square error (MMSE)‐based techniques are often used for the joint iterative detection and decoding that is for coded multi‐input–multi‐output (MIMO) system due to a sound complexity and the performance trade‐off. This study proposes an enhanced MMSE‐based soft MIMO‐detection scheme by using three main ideas. The first idea is an efficient complexity‐reduced soft‐bit estimation technique, the second one is a performance improvement method utilised inside the MMSE detection process, and the third one is a complexity‐reduced soft‐symbol estimation method for quadrature amplitude modulation. The proposed ideas enable the interference‐cancellation processes to be activated in parallel on each symbol layer, thereby reducing the processing time. The simulation results show that the proposed method efficiently contributes to the improvement of the performance in addition to its reduction of the linear‐order complexity.

In this study, the authors propose a joint iterative detection and decoding (JIDD) method for a turbo coded multiple‐input multiple‐output (MIMO) system, with a linear order of complexity. Accurate estimation of soft information should be conditioned for excellent performance of the JIDD, but it usually requires an exponential order of complexity. They propose a method which improves the performance of soft interference cancellation minimum mean‐squared error (SIC‐MMSE) method by increasing the reliability of the soft information utilised for interference cancellation. The proposed method utilises a posteriori probabilities from the MIMO detector as well as a priori probabilities from the turbo decoder, and perform soft minimum mean‐squared error filtering for symbol level detection. With this approach, soft information is fully fed into the symbol detection process, and thus the reliability of soft symbol is increased. In addition, they can separate out the bit‐level soft estimation process by using a simple linear method. Simulation results show that the proposed method provides substantial complexity reduction, with a bit error rate performance comparable to the conventional SIC‐MMSE methods.

List‐sphere detection (LSD) is a sub‐optimal multiple‐input multiple‐output (MIMO) detection scheme which searches candidate symbol vectors that lie within a sphere of a given radius. This study presents an efficient LSD based method for a joint iterative MIMO detection scheme. The proposed method utilises a channel condition in order to define the list size. During the search process, the radius is adaptively updated to reduce the computational complexity. Owing to the list size and corresponding radius are adaptively determined by the channel condition, the authors can operate the detector at the most appropriate complexity to produce the required performance. Simulation results show that the proposed methods provide substantial complexity reduction without bit error rate performance degradation.