Recent 3D-aware GANs rely on volumetric rendering techniques to disentangle the pose and appearance of objects, de facto generating entire 3D volumes rather than single-view 2D images from a latent code. Complex image editing tasks can be performed in standard 2D-based GANs (e.g., StyleGAN models) as manipulation of latent dimensions. However, to the best of our knowledge, similar properties have only been partially explored for 3D-aware GAN models. This work aims to fill this gap by showing the limitations of existing methods and proposing LatentSwap3D, a model-agnostic approach designed to enable attribute editing in the latent space of pre-trained 3D-aware GANs. We first identify the most relevant dimensions in the latent space of the model controlling the targeted attribute by relying on the feature importance ranking of a random forest classifier. Then, to apply the transformation, we swap the top-K most relevant latent dimensions of the image being edited with an image exhibiting the desired attribute. Despite its simplicity, LatentSwap3D provides remarkable semantic edits in a disentangled manner and outperforms alternative approaches both qualitatively and quantitatively. We demonstrate our semantic edit approach on various 3D-aware generative models such as pi-GAN, GIRAFFE, StyleSDF, MVCGAN, EG3D and VolumeGAN, and on diverse datasets, such as FFHQ, AFHQ, Cats, MetFaces, and CompCars. The project page can be found: \url{https://enisimsar.github.io/latentswap3d/}.
Current methods for image-to-image translation produce compelling results, however, the applied transformation is difficult to control, since existing mechanisms are often limited and non-intuitive. We propose ParGAN, a generalization of the cycle-consistent GAN framework to learn image transformations with simple and intuitive controls. The proposed generator takes as input both an image and a parametrization of the transformation. We train this network to preserve the content of the input image while ensuring that the result is consistent with the given parametrization. Our approach does not require paired data and can learn transformations across several tasks and datasets. We show how, with disjoint image domains with no annotated parametrization, our framework can create smooth interpolations as well as learn multiple transformations simultaneously.
Most modern deep learning-based multi-view 3D reconstruction techniques use RNNs or fusion modules to combine information from multiple images after encoding them. These two separate steps have loose connections and do not consider all available information while encoding each view. We propose LegoFormer, a transformer-based model that unifies object reconstruction under a single framework and parametrizes the reconstructed occupancy grid by its decomposition factors. This reformulation allows the prediction of an object as a set of independent structures then aggregated to obtain the final reconstruction. Experiments conducted on ShapeNet display the competitive performance of our network with respect to the state-of-the-art methods. We also demonstrate how the use of self-attention leads to increased interpretability of the model output.
Novel view synthesis from a single image aims at generating novel views from a single input image of an object. Several works recently achieved remarkable results, though require some form of multi-view supervision at training time, therefore limiting their deployment in real scenarios. This work aims at relaxing this assumption enabling training of conditional generative model for novel view synthesis in a completely unsupervised manner. We first pre-train a purely generative decoder model using a GAN formulation while at the same time training an encoder network to invert the mapping from latent code to images. Then we swap encoder and decoder and train the network as a conditioned GAN with a mixture of auto-encoder-like objective and self-distillation. At test time, given a view of an object, our model first embeds the image content in a latent code and regresses its pose w.r.t. a canonical reference system, then generates novel views of it by keeping the code and varying the pose. We show that our framework achieves results comparable to the state of the art on ShapeNet and that it can be employed on unconstrained collections of natural images, where no competing method can be trained.
Domain generalization aims at training machine learning models to perform robustly across different and unseen domains. Several recent methods use multiple datasets to train models to extract domain-invariant features, hoping to generalize to unseen domains. Instead, first we explicitly train domain-dependant representations by using ad-hoc batch normalization layers to collect independent domain's statistics. Then, we propose to use these statistics to map domains in a shared latent space, where membership to a domain can be measured by means of a distance function. At test time, we project samples from an unknown domain into the same space and infer properties of their domain as a linear combination of the known ones. We apply the same mapping strategy at training and test time, learning both a latent representation and a powerful but lightweight ensemble model. We show a significant increase in classification accuracy over current state-of-the-art techniques on popular domain generalization benchmarks: PACS, Office-31 and Office-Caltech.
Estimating the 3D shape of an object from a single or multiple images has gained popularity thanks to the recent breakthroughs powered by deep learning. Most approaches regress the full object shape in a canonical pose, possibly extrapolating the occluded parts based on the learned priors. However, their viewpoint invariant technique often discards the unique structures visible from the input images. In contrast, this paper proposes to rely on viewpoint variant reconstructions by merging the visible information from the given views. Our approach is divided into three steps. Starting from the sparse views of the object, we first align them into a common coordinate system by estimating the relative pose between all the pairs. Then, inspired by the traditional voxel carving, we generate an occupancy grid of the object taken from the silhouette on the images and their relative poses. Finally, we refine the initial reconstruction to build a clean 3D model which preserves the details from each viewpoint. To validate the proposed method, we perform a comprehensive evaluation on the ShapeNet reference benchmark in terms of relative pose estimation and 3D shape reconstruction.
Depth estimation from stereo images is carried out with unmatched results by convolutional neural networks trained end-to-end to regress dense disparities. Like for most tasks, this is possible if large amounts of labelled samples are available for training, possibly covering the whole data distribution encountered at deployment time. Being such an assumption systematically unmet in real applications, the capacity of adapting to any unseen setting becomes of paramount importance. Purposely, we propose a continual adaptation paradigm for deep stereo networks designed to deal with challenging and ever-changing environments. We design a lightweight and modular architecture, Modularly ADaptive Network (MADNet), and formulate Modular ADaptation algorithms (MAD, MAD++) which permit efficient optimization of independent sub-portions of the entire network. In our paradigm, the learning signals needed to continuously adapt models online can be sourced from self-supervision via right-to-left image warping or from traditional stereo algorithms. With both sources, no other data than the input images being gathered at deployment time are needed. Thus, our network architecture and adaptation algorithms realize the first real-time self-adaptive deep stereo system and pave the way for a new paradigm that can facilitate practical deployment of end-to-end architectures for dense disparity regression.
State-of-the-art approaches to infer dense depth measurements from images rely on CNNs trained end-to-end on a vast amount of data. However, these approaches suffer a drastic drop in accuracy when dealing with environments much different in appearance and/or context from those observed at training time. This domain shift issue is usually addressed by fine-tuning on smaller sets of images from the target domain annotated with depth labels. Unfortunately, relying on such supervised labeling is seldom feasible in most practical settings. Therefore, we propose an unsupervised domain adaptation technique which does not require groundtruth labels. Our method relies only on image pairs and leverages on classical stereo algorithms to produce disparity measurements alongside with confidence estimators to assess upon their reliability. We propose to fine-tune both depth-from-stereo as well as depth-from-mono architectures by a novel confidence-guided loss function that handles the measured disparities as noisy labels weighted according to the estimated confidence. Extensive experimental results based on standard datasets and evaluation protocols prove that our technique can address effectively the domain shift issue with both stereo and monocular depth prediction architectures and outperforms other state-of-the-art unsupervised loss functions that may be alternatively deployed to pursue domain adaptation.
In this paper, we propose Augmented Reality Semi-automatic labeling (ARS), a semi-automatic method which leverages on moving a 2D camera by means of a robot, proving precise camera tracking, and an augmented reality pen to define initial object bounding box, to create large labeled datasets with minimal human intervention. By removing the burden of generating annotated data from humans, we make the Deep Learning technique applied to computer vision, that typically requires very large datasets, truly automated and reliable. With the ARS pipeline, we created effortlessly two novel datasets, one on electromechanical components (industrial scenario) and one on fruits (daily-living scenario), and trained robustly two state-of-the-art object detectors, based on convolutional neural networks, such as YOLO and SSD. With respect to the conventional manual annotation of 1000 frames that takes us slightly more than 10 hours, the proposed approach based on ARS allows annotating 9 sequences of about 35000 frames in less than one hour, with a gain factor of about 450. Moreover, both the precision and recall of object detection is increased by about 15\% with respect to manual labeling. All our software is available as a ROS package in a public repository alongside the novel annotated datasets.