Current 3D object detectors for autonomous driving are almost entirely trained on human-annotated data. Although of high quality, the generation of such data is laborious and costly, restricting them to a few specific locations and object types. This paper proposes an alternative approach entirely based on unlabeled data, which can be collected cheaply and in abundance almost everywhere on earth. Our ap- proach leverages several simple common sense heuristics to create an initial set of approximate seed labels. For ex- ample, relevant traffic participants are generally not per- sistent across multiple traversals of the same route, do not fly, and are never under ground. We demonstrate that these seed labels are highly effective to bootstrap a surpris- ingly accurate detector through repeated self-training with- out a single human annotated label. Code is available at https:// github.com/ YurongYou/ MODEST .
more »
« less
Learning to Detect Mobile Objects from LiDAR Scans Without Labels
Current 3D object detectors for autonomous driving are almost entirely trained on human-annotated data. Although of high quality, the generation of such data is laborious and costly, restricting them to a few specific locations and object types. This paper proposes an alternative approach entirely based on unlabeled data, which can be collected cheaply and in abundance almost everywhere on earth. Our approach leverages several simple common sense heuristics to create an initial set of approximate seed labels. For example, relevant traffic participants are generally not persistent across multiple traversals of the same route, do not fly, and are never under ground. We demonstrate that these seed labels are highly effective to bootstrap a surprisingly accurate detector through repeated self-training without a single human annotated label. Code is available at https://github.com/YurongYou/MODEST.
more »
« less
- Award ID(s):
- 2118240
- NSF-PAR ID:
- 10338440
- Date Published:
- Journal Name:
- IEEE / CVF Computer Vision and Pattern Recognition Conference
- Page Range / eLocation ID:
- 1130 - 1140
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Background: When phenotypic characters are described in the literature, they may be constrained or clarified with additional information such as the location or degree of expression, these terms are called “modifiers”. With effort underway to convert narrative character descriptions to computable data, ontologies for such modifiers are needed. Such ontologies can also be used to guide term usage in future publications. Spatial and method modifiers are the subjects of ontologies that already have been developed or are under development. In this work, frequency (e.g., rarely, usually), certainty (e.g., probably, definitely), degree (e.g., slightly, extremely), and coverage modifiers (e.g., sparsely, entirely) are collected, reviewed, and used to create two modifier ontologies with different design considerations. The basic goal is to express the sequential relationships within a type of modifiers, for example, usually is more frequent than rarely, in order to allow data annotated with ontology terms to be classified accordingly. Method: Two designs are proposed for the ontology, both using the list pattern: a closed ordered list (i.e., five-bin design) and an open ordered list design. The five-bin design puts the modifier terms into a set of 5 fixed bins with interval object properties, for example, one_level_more/less_frequently_than, where new terms can only be added as synonyms to existing classes. The open list approach starts with 5 bins, but supports the extensibility of the list via ordinal properties, for example, more/less_frequently_than, allowing new terms to be inserted as a new class anywhere in the list. The consequences of the different design decisions are discussed in the paper. CharaParser was used to extract modifiers from plant, ant, and other taxonomic descriptions. After a manual screening, 130 modifier words were selected as the candidate terms for the modifier ontologies. Four curators/experts (three biologists and one information scientist specialized in biosemantics) reviewed and categorized the terms into 20 bins using the Ontology Term Organizer (OTO) (http://biosemantics.arizona.edu/OTO). Inter-curator variations were reviewed and expressed in the final ontologies. Results: Frequency, certainty, degree, and coverage terms with complete agreement among all curators were used as class labels or exact synonyms. Terms with different interpretations were either excluded or included using “broader synonym” or “not recommended” annotation properties. These annotations explicitly allow for the user to be aware of the semantic ambiguity associated with the terms and whether they should be used with caution or avoided. Expert categorization results showed that 16 out of 20 bins contained terms with full agreements, suggesting differentiating the modifiers into 5 levels/bins balances the need to differentiate modifiers and the need for the ontology to reflect user consensus. Two ontologies, developed using the Protege ontology editor, are made available as OWL files and can be downloaded from https://github.com/biosemantics/ontologies. Contribution: We built the first two modifier ontologies following a consensus-based approach with terms commonly used in taxonomic literature. The five-bin ontology has been used in the Explorer of Taxon Concepts web toolkit to compute the similarity between characters extracted from literature to facilitate taxon concepts alignments. The two ontologies will also be used in an ontology-informed authoring tool for taxonomists to facilitate consistency in modifier term usage.more » « less
-
null (Ed.)The success of supervised learning requires large-scale ground truth labels which are very expensive, time- consuming, or may need special skills to annotate. To address this issue, many self- or un-supervised methods are developed. Unlike most existing self-supervised methods to learn only 2D image features or only 3D point cloud features, this paper presents a novel and effective self-supervised learning approach to jointly learn both 2D image features and 3D point cloud features by exploiting cross-modality and cross-view correspondences without using any human annotated labels. Specifically, 2D image features of rendered images from different views are extracted by a 2D convolutional neural network, and 3D point cloud features are extracted by a graph convolution neural network. Two types of features are fed into a two-layer fully connected neural network to estimate the cross-modality correspondence. The three networks are jointly trained (i.e. cross-modality) by verifying whether two sampled data of different modalities belong to the same object, meanwhile, the 2D convolutional neural network is additionally optimized through minimizing intra-object distance while maximizing inter-object distance of rendered images in different views (i.e. cross-view). The effectiveness of the learned 2D and 3D features is evaluated by transferring them on five different tasks including multi-view 2D shape recognition, 3D shape recognition, multi-view 2D shape retrieval, 3D shape retrieval, and 3D part-segmentation. Extensive evaluations on all the five different tasks across different datasets demonstrate strong generalization and effectiveness of the learned 2D and 3D features by the proposed self-supervised method.more » « less
-
Meila, Marina ; Zhang, Tong (Ed.)The label noise transition matrix, characterizing the probabilities of a training instance being wrongly annotated, is crucial to designing popular solutions to learning with noisy labels. Existing works heavily rely on finding “anchor points” or their approximates, defined as instances belonging to a particular class almost surely. Nonetheless, finding anchor points remains a non-trivial task, and the estimation accuracy is also often throttled by the number of available anchor points. In this paper, we propose an alternative option to the above task. Our main contribution is the discovery of an efficient estimation procedure based on a clusterability condition. We prove that with clusterable representations of features, using up to third-order consensuses of noisy labels among neighbor representations is sufficient to estimate a unique transition matrix. Compared with methods using anchor points, our approach uses substantially more instances and benefits from a much better sample complexity. We demonstrate the estimation accuracy and advantages of our estimates using both synthetic noisy labels (on CIFAR-10/100) and real human-level noisy labels (on Clothing1M and our self-collected human-annotated CIFAR-10). Our code and human-level noisy CIFAR-10 labels are available at https://github.com/UCSC-REAL/HOC.more » « less
-
more » « less
Abstract Advances in visual perceptual tasks have been mainly driven by the amount, and types, of annotations of large-scale datasets. Researchers have focused on fully-supervised settings to train models using offline epoch-based schemes. Despite the evident advancements, limitations and cost of manually annotated datasets have hindered further development for event perceptual tasks, such as detection and localization of objects and events in videos. The problem is more apparent in zoological applications due to the scarcity of annotations and length of videos-most videos are at most ten minutes long. Inspired by cognitive theories, we present a self-supervised perceptual prediction framework to tackle the problem of temporal event segmentation by building a stable representation of event-related objects. The approach is simple but effective. We rely on LSTM predictions of high-level features computed by a standard deep learning backbone. For spatial segmentation, the stable representation of the object is used by an attention mechanism to filter the input features before the prediction step. The self-learned attention maps effectively localize the object as a side effect of perceptual prediction. We demonstrate our approach on long videos from continuous wildlife video monitoring, spanning multiple days at 25 FPS. We aim to facilitate automated ethogramming by detecting and localizing events without the need for labels. Our approach is trained in an online manner on streaming input and requires only a single pass through the video, with no separate training set. Given the lack of long and realistic (includes real-world challenges) datasets, we introduce a new wildlife video dataset–nest monitoring of the Kagu (a flightless bird from New Caledonia)–to benchmark our approach. Our dataset features a video from 10 days (over 23 million frames) of continuous monitoring of the Kagu in its natural habitat. We annotate every frame with bounding boxes and event labels. Additionally, each frame is annotated with time-of-day and illumination conditions. We will make the dataset, which is the first of its kind, and the code available to the research community. We find that the approach significantly outperforms other self-supervised, traditional (e.g., Optical Flow, Background Subtraction) and NN-based (e.g., PA-DPC, DINO, iBOT), baselines and performs on par with supervised boundary detection approaches (i.e., PC). At a recall rate of 80%, our best performing model detects one false positive activity every 50 min of training. On average, we at least double the performance of self-supervised approaches for spatial segmentation. Additionally, we show that our approach is robust to various environmental conditions (e.g., moving shadows). We also benchmark the framework on other datasets (i.e., Kinetics-GEBD, TAPOS) from different domains to demonstrate its generalizability. The data and code are available on our project page:
https://aix.eng.usf.edu/research_automated_ethogramming.html
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- The DOI is not currently available.
