Computational Baby Learning

By Xiaodan Liang, Si Liu, Yunchao Wei, Luoqi Liu, Liang Lin, Shuicheng Yan. 2014

26 citations recorded by the source corpus.

Intuitive observations show that a baby may inherently possess the capability of recognizing a new visual concept (e.g., chair, dog) by learning from only very few positive instances taught by parent(s) or others, and this recognition capability can be gradually further improved by exploring and/or interacting with the real instances in the physical world. Inspired by these observations, we propose a computational model for slightly-supervised object detection, based on prior knowledge modelling, exemplar learning and learning with video contexts. The prior knowledge is modeled with a pre-trained Convolutional Neural Network (CNN). When very few instances of a new concept are given, an initial concept detector is built by exemplar learning over the deep features from the pre-trained CNN. Simulating the baby's interaction with physical world, the well-designed tracking solution is then used to discover more diverse instances from the massive online unlabeled videos. Once a positive instance is detected/identified with high score in each video, more variable instances possibly from different view-angles and/or different distances are tracked and accumulated. Then the concept detector can be fine-tuned based on these new instances. This process can be repeated again and again till we obtain a very mature concept detector. Extensive experiments on Pascal VOC-07/10/12 object detection datasets well demonstrate the effectiveness of our framework. It can beat the state-of-the-art full-training based performances by learning from very few samples for each object category, along with about 20,000 unlabeled videos.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Computational Linguistics & Semantics, Cognitive Science & Psychology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Computational Linguistics & Semantics

In Cognitive Science & Psychology

Citations in the corpus

Object Detection Networks on Convolutional Feature Maps (2015, neutral): “7 64.3 43.9 80.4 71.6 74.0 60.0 30.8 63.4 52.0 63.5 58.7 R-CNN, bb [9] VOC 12 62.4 79.6 72.7 61.9 41.2 41.9 65.9 66.4 84.6 38.5 67.2 46.7 82.0 74.8 76.0 65.2 35.6 65.4 54.2 67.4 60.3 BabyLearning, bb [20] VOC+ext. 63.2 78.0 74.2 61.3 45.7 42.7 68.2 66.8 80.2 40.6 70.0 49.8 79.0 74.5 77.9 64.0 35.3 6”
Learning to See by Moving (2015, methodology): “Our best performing architecture (C96-P-C256-P), outperforms convolutional deep belief networks [27], a previous approach based on learning features invariant to transformations [34] and features learnt by the default autoencoder network provided with Caffe[21].”
BoxSup: Exploiting Bounding Boxes to Supervise Convolutional Networks for Semantic Segmentation (2015, methodology): “In [17], the training set for object detection is augmented by visual tracking results obtained from videos.”
Unsupervised Learning of Visual Representations Using Videos (2015, methodology): “[30] used thousands of keywords to retrieve videos from YouTube.”
Proposal-Free Network for Instance-Level Object Segmentation (2015, neutral): “Over the past few decades, two of the most popular object recognition tasks, object detection and semantic segmentation, have received a lot of attention.”
Transitive Invariance for Self-Supervised Visual Representation Learning (2017, methodology): “To prepare the data for training, we download the 100K videos from YouTube using the URLs provided by [33, 56].”
Watch and learn: Semi-supervised learning of object detectors from videos (2015, methodology): “A concurrent work [33] utilizes weakly labeled video for improving detection.”
Adaptive Object Detection Using Adjacency and Zoom Prediction (2015, background): “While these approaches are computationally efficient, they rely on fixed image regions as anchors for predictions.”
Revisiting Dilated Convolution: A Simple Approach for Weakly- and Semi-Supervised Semantic Segmentation (2018, background): “Weakly-supervised image recognition approaches [11, 15, 16, 19, 23, 38–42 , 44] Recently, some top-down approaches [43,46] propose to between the state-of-the-art CAM [46] (the 2nd row) and ours (the last row) on quality of the produced object localization maps.”
Track and Transfer: Watching Videos to Simulate Strong Human Supervision for Weakly-Supervised Object Detection (2016, methodology): “Finally, recent work uses videos for semi-supervised object detection with bounding box annotations as initialization [21], or trains a CNN for feature learning using tracking as supervision and fine-tuning the learned representation with bounding box annotations for detection [38].”
STC: A Simple to Complex Framework for Weakly-Supervised Semantic Segmentation (2015, background): “mework ﬁrst learns from simple easy images and then applies the learned network to complex ones, which is also related to self-paced learning [27]. Recently, various computer vision applications [28]–[30] based on self-paced learning have been proposed. In speciﬁc, Tang et al. [28] adapted object de”
Temporal Localization of Fine-Grained Actions in Videos by Domain Transfer from Web Images (2015, background): “Liang et al. [19] propose a computational baby learning framework.”
Understanding Blooming Human Groups in Social Networks (2015, background): “[23] proposed a framework to learning the new labels with few samples by exploring the positive instances by exemplar learning.”
Unsupervised Joint Mining of Deep Features and Image Labels for Large-Scale Radiology Image Categorization and Scene Recognition (2017, background): “ImageNet pre-trained deep CNN models [26, 30, 37] serve an indispensable role to be bootstrapped or fine-tuned [24] for all externallysourced data exploitation tasks [36, 5].”
Scale-Aware Fast R-CNN for Pedestrian Detection (2015, methodology): “…offer different category-level conﬁ-dence scores and bounding box regressions for each proposal, which are then combined to generate the ﬁnal detection results using two scale-aware weights predicted by a scale-aware weighting layer that performs a gate function deﬁned over the proposal size.”