What Is Holding Back Convnets for Detection?

By Bojan Pepik, Rodrigo Benenson, Tobias Ritschel, Bernt Schiele. 2015

66 citations recorded by the source corpus.

Convolutional neural networks have recently shown excellent results in general object detection and many other tasks. Albeit very effective, they involve many user-defined design choices. In this paper we want to better understand these choices by inspecting two key aspects “what did the network learn?”, and “what can the network learn?”. We exploit new annotations (Pascal3D+), to enable a new empirical analysis of the R-CNN detector. Despite common belief, our results indicate that existing state-of-the-art convnets are not invariant to various appearance factors. In fact, all considered networks have similar weak points which cannot be mitigated by simply increasing the training data (architectural changes are needed). We show that overall performance can improve when using image renderings as data augmentation. We report the best known results on Pascal3D+ detection and view-point estimation tasks.

Subfields: Computer Vision and Pattern Recognition

How researchers see this paper

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

In Computer Vision and Pattern Recognition

Citations in the corpus

DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks (2015, methodology): “Similarly to the binary case, we first present the proposed approach for the linear case and then we generalize it to other classifiers.”
Marr Revisited: 2D-3D Alignment via Surface Normal Prediction (2016, methodology): “In this paper, we propose to bring back the ideas put forth by Marr and develop a computational framework for extracting 2.5D representation followed by 3D volumetric estimation.”
Synthetic Data for Deep Learning (2019, background): “[447], and Aubry and Russell [30] used synthetic data to study the properties of deep convolutional networks, in particular robustness to various transformations, since synthetic data is easy to manipulate in a predefined way.”
On the Robustness of Semantic Segmentation Models to Adversarial Attacks (2017, background): “This is because CNNs are not invariant to scale, and a range of other transformations [33, 75, 45].”
Deep Exemplar 2D-3D Detection by Adapting from Real to Rendered Views (2015, methodology): “We applied our method to two tasks: instance detection, where we evaluated on the IKEA dataset [36], and object category detection, where we out-perform Aubry et al. [3] for “chair” detection on a subset of the Pascal VOC dataset.”
Deep Learning for Generic Object Detection: A Survey (2018, background): “These modalities raise new challenges in effectively using depth (Chen et al. 2015c; Pepik et al. 2015; Xiang et al. 2014; Wu et al. 2015), video (Feichtenhofer et al.”