Victor S. Lempitsky

The devil is in the details: an evaluation of recent feature encoding methods

A large number of novel encodings for bag of visual words models have been proposed in the past two years to improve on the standard histogram of quantized local features. Examples include locality-constrained linear encoding [23], improved Fisher encoding [17], super vector encoding [27], and kernel codebook encoding [20]. While several authors have reported very good results on the challenging PASCAL VOC classification data by means of these new techniques, differences in the feature computation and learning algorithms, missing details in the description of the methods, and different tuning of the various components, make it impossible to compare directly these methods and hard to reproduce the results reported. This paper addresses these shortcomings by carrying out a rigorous evaluation of these new techniques by: (1) fixing the other elements of the pipeline (features, learning, tuning); (2) disclosing all the implementation details, and (3) identifying both those aspects of each method which are particularly important to achieve good performance, and those aspects which are less critical. This allows a consistent comparative analysis of these encoding methods. Several conclusions drawn from our analysis cannot be inferred from the original publications.

Domain-adversarial training of neural networks

We introduce a new representation learning approach for domain adaptation, in which data at training and test time come from similar but different distributions. Our approach is directly inspired by the theory on domain adaptation suggesting that, for effective domain transfer to be achieved, predictions must be made based on features that cannot discriminate between the training (source) and test (target) domains. The approach implements this idea in the context of neural network architectures that are trained on labeled data from the source domain and unlabeled data from the target domain (no labeled target-domain data is necessary). As the training progresses, the approach promotes the emergence of features that are (i) discriminative for the main learning task on the source domain and (ii) indiscriminate with respect to the shift between the domains. We show that this adaptation behaviour can be achieved in almost any feed-forward model by augmenting it with few standard layers and a new gradient reversal layer. The resulting augmented architecture can be trained using standard backpropagation and stochastic gradient descent, and can thus be implemented with little effort using any of the deep learning packages. We demonstrate the success of our approach for two distinct classification problems (document sentiment analysis and image classification), where state-of-the-art domain adaptation performance on standard benchmarks is achieved. We also validate the approach for descriptor learning task in the context of person re-identification application.

Hough Forests for Object Detection, Tracking, and Action Recognition

The paper introduces Hough forests, which are random forests adapted to perform a generalized Hough transform in an efficient way. Compared to previous Hough-based systems such as implicit shape models, Hough forests improve the performance of the generalized Hough transform for object detection on a categorical level. At the same time, their flexibility permits extensions of the Hough transform to new domains such as object tracking and action recognition. Hough forests can be regarded as task-adapted codebooks of local appearance that allow fast supervised training and fast matching at test time. They achieve high detection accuracy since the entries of such codebooks are optimized to cast Hough votes with small variance and since their efficiency permits dense sampling of local image patches or video cuboids during detection. The efficacy of Hough forests for a set of computer vision tasks is validated through experiments on a large set of publicly available benchmark data sets and comparisons with the state-of-the-art.

Neural Codes for Image Retrieval

It has been shown that the activations invoked by an image within the top layers of a large convolutional neural network provide a high-level descriptor of the visual content of the image. In this paper, we investigate the use of such descriptors (neural codes) within the image retrieval application. In the experiments with several standard retrieval benchmarks, we establish that neural codes perform competitively even when the convolutional neural network has been trained for an unrelated classification task (e.g. Image-Net). We also evaluate the improvement in the retrieval performance of neural codes, when the network is retrained on a dataset of images that are similar to images encountered at test time.

Class-specific Hough forests for object detection

We present a method for the detection of instances of an object class, such as cars or pedestrians, in natural images. Similarly to some previous works, this is accomplished via generalized Hough transform, where the detections of individual object parts cast probabilistic votes for possible locations of the centroid of the whole object; the detection hypotheses then correspond to the maxima of the Hough image that accumulates the votes from all parts. However, whereas the previous methods detect object parts using generative codebooks of part appearances, we take a more discriminative approach to object part detection. Towards this end, we train a class-specific Hough forest, which is a random forest that directly maps the image patch appearance to the probabilistic vote about the possible location of the object centroid. We demonstrate that Hough forests improve the results of the Hough-transform object detection significantly and achieve state-of-the-art performance for several classes and datasets.

Image segmentation with a bounding box prior

User-provided object bounding box is a simple and popular interaction paradigm considered by many existing interactive image segmentation frameworks. However, these frameworks tend to exploit the provided bounding box merely to exclude its exterior from consideration and sometimes to initialize the energy minimization. In this paper, we discuss how the bounding box can be further used to impose a powerful topological prior, which prevents the solution from excessive shrinking and ensures that the user-provided box bounds the segmentation in a sufficiently tight way. The prior is expressed using hard constraints incorporated into the global energy minimization framework leading to an NP-hard integer program. We then investigate the possible optimization strategies including linear relaxation as well as a new graph cut algorithm called pinpointing. The latter can be used either as a rounding method for the fractional LP solution, which is provably better than thresholding-based rounding, or as a fast standalone heuristic. We evaluate the proposed algorithms on a publicly available dataset, and demonstrate the practical benefits of the new prior both qualitatively and quantitatively.

Fusion Moves for Markov Random Field Optimization

The efficient application of graph cuts to Markov Random Fields (MRFs) with multiple discrete or continuous labels remains an open question. In this paper, we demonstrate one possible way of achieving this by using graph cuts to combine pairs of suboptimal labelings or solutions. We call this combination process the fusion move. By employing recently developed graph-cut-based algorithms (so-called QPBO-graph cut), the fusion move can efficiently combine two proposal labelings in a theoretically sound way, which is in practice often globally optimal. We demonstrate that fusion moves generalize many previous graph-cut approaches, which allows them to be used as building blocks within a broader variety of optimization schemes than were considered before. In particular, we propose new optimization schemes for computer vision MRFs with applications to image restoration, stereo, and optical flow, among others. Within these schemes the fusion moves are used 1) for the parallelization of MRF optimization into several threads, 2) for fast MRF optimization by combining cheap-to-compute solutions, and 3) for the optimization of highly nonconvex continuous-labeled MRFs with 2D labels. Our final example is a nonvision MRF concerned with cartographic label placement, where fusion moves can be used to improve the performance of a standard inference method (loopy belief propagation).

On Detection of Multiple Object Instances Using Hough Transforms

Hough transform-based methods for detecting multiple objects use nonmaxima suppression or mode seeking to locate and distinguish peaks in Hough images. Such postprocessing requires the tuning of many parameters and is often fragile, especially when objects are located spatially close to each other. In this paper, we develop a new probabilistic framework for object detection which is related to the Hough transform. It shares the simplicity and wide applicability of the Hough transform but, at the same time, bypasses the problem of multiple peak identification in Hough images and permits detection of multiple objects without invoking nonmaximum suppression heuristics. Our experiments demonstrate that this method results in a significant improvement in detection accuracy both for the classical task of straight line detection and for a more modern category-level (pedestrian) detection problem.

Aggregating Local Deep Features for Image Retrieval

Several recent works have shown that image descriptors produced by deep convolutional neural networks provide state-of-the-art performance for image classification and retrieval problems. It also has been shown that the activations from the convolutional layers can be interpreted as local features describing particular image regions. These local features can be aggregated using aggregating methods developed for local features (e.g. Fisher vectors), thus providing new powerful global descriptor. In this paper we investigate possible ways to aggregate local deep features to produce compact descriptors for image retrieval. First, we show that deep features and traditional hand-engineered features have quite different distributions of pairwise similarities, hence existing aggregation methods have to be carefully re-evaluated. Such re-evaluation reveals that in contrast to shallow features, the simple aggregation method based on sum pooling provides the best performance for deep convolutional features. This method is efficient, has few parameters, and bears little risk of overfitting when e.g. learning the PCA matrix. In addition, we suggest a simple yet efficient query expansion scheme suitable for the proposed aggregation method. Overall, the new compact global descriptor improves the state-of-the-art on four common benchmarks considerably.

The inverted multi-index

A new data structure for efficient similarity search in very large dataseis of high-dimensional vectors is introduced. This structure called the inverted multi-index generalizes the inverted index idea by replacing the standard quantization within inverted indices with product quantization. For very similar retrieval complexity and preprocessing time, inverted multi-indices achieve a much denser subdivision of the search space compared to inverted indices, while retaining their memory efficiency. Our experiments with large dataseis of SIFT and GIST vectors demonstrate that because of the denser subdivision, inverted multi-indices are able to return much shorter candidate lists with higher recall. Augmented with a suitable reranking procedure, multi-indices were able to improve the speed of approximate nearest neighbor search on the dataset of 1 billion SIFT vectors by an order of magnitude compared to the best previously published systems, while achieving better recall and incurring only few percent of memory overhead.