Classifier guidance -- using the gradients of an image classifier to steer the generations of a diffusion model -- has the potential to dramatically expand the creative control over image generation and editing. However, currently classifier guidance requires either training new noise-aware models to obtain accurate gradients or using a one-step denoising approximation of the final generation, which leads to misaligned gradients and sub-optimal control. We highlight this approximation's shortcomings and propose a novel guidance method: Direct Optimization of Diffusion Latents (DOODL), which enables plug-and-play guidance by optimizing diffusion latents w.r.t. the gradients of a pre-trained classifier on the true generated pixels, using an invertible diffusion process to achieve memory-efficient backpropagation. Showcasing the potential of more precise guidance, DOODL outperforms one-step classifier guidance on computational and human evaluation metrics across different forms of guidance: using CLIP guidance to improve generations of complex prompts from DrawBench, using fine-grained visual classifiers to expand the vocabulary of Stable Diffusion, enabling image-conditioned generation with a CLIP visual encoder, and improving image aesthetics using an aesthetic scoring network.
The incredible generative ability of large-scale text-to-image (T2I) models has demonstrated strong power of learning complex structures and meaningful semantics. However, relying solely on text prompts cannot fully take advantage of the knowledge learned by the model, especially when flexible and accurate structure control is needed. In this paper, we aim to ``dig out" the capabilities that T2I models have implicitly learned, and then explicitly use them to control the generation more granularly. Specifically, we propose to learn simple and small T2I-Adapters to align internal knowledge in T2I models with external control signals, while freezing the original large T2I models. In this way, we can train various adapters according to different conditions, and achieve rich control and editing effects. Further, the proposed T2I-Adapters have attractive properties of practical value, such as composability and generalization ability. Extensive experiments demonstrate that our T2I-Adapter has promising generation quality and a wide range of applications.
We present a simple, efficient, and scalable unfolding network, SAUNet, to simplify the network design with an adaptive alternate optimization framework for hyperspectral image (HSI) reconstruction. SAUNet customizes a Residual Adaptive ADMM Framework (R2ADMM) to connect each stage of the network via a group of learnable parameters to promote the usage of mask prior, which greatly stabilizes training and solves the accuracy degradation issue. Additionally, we introduce a simple convolutional modulation block (CMB), which leads to efficient training, easy scale-up, and less computation. Coupling these two designs, SAUNet can be scaled to non-trivial 13 stages with continuous improvement. Without bells and whistles, SAUNet improves both performance and speed compared with the previous state-of-the-art counterparts, which makes it feasible for practical high-resolution HSI reconstruction scenarios. We set new records on CAVE and KAIST HSI reconstruction benchmarks. Code and models are available at https://github.com/hustvl/SAUNet.
Salient object detection (SOD) is a task that involves identifying and segmenting the most visually prominent object in an image. Existing solutions can accomplish this use a multi-scale feature fusion mechanism to detect the global context of an image. However, as there is no consideration of the structures in the image nor the relations between distant pixels, conventional methods cannot deal with complex scenes effectively. In this paper, we propose an adaptive graph convolution module (AGCM) to overcome these limitations. Prototype features are initially extracted from the input image using a learnable region generation layer that spatially groups features in the image. The prototype features are then refined by propagating information between them based on a graph architecture, where each feature is regarded as a node. Experimental results show that the proposed AGCM dramatically improves the SOD performance both quantitatively and quantitatively.
In this paper, we present a comprehensive study on semantic segmentation with the Pascal VOC dataset. Here, we have to label each pixel with a class which in turn segments the entire image based on the objects/entities present. To tackle this, we firstly use a Fully Convolution Network (FCN) baseline which gave 71.31% pixel accuracy and 0.0527 mean IoU. We analyze its performance and working and subsequently address the issues in the baseline with three improvements: a) cosine annealing learning rate scheduler(pixel accuracy: 72.86%, IoU: 0.0529), b) data augmentation(pixel accuracy: 69.88%, IoU: 0.0585) c) class imbalance weights(pixel accuracy: 68.98%, IoU: 0.0596). Apart from these changes in training pipeline, we also explore three different architectures: a) Our proposed model -- Advanced FCN (pixel accuracy: 67.20%, IoU: 0.0602) b) Transfer Learning with ResNet (Best performance) (pixel accuracy: 71.33%, IoU: 0.0926 ) c) U-Net(pixel accuracy: 72.15%, IoU: 0.0649). We observe that the improvements help in greatly improving the performance, as reflected both, in metrics and segmentation maps. Interestingly, we observe that among the improvements, dataset augmentation has the greatest contribution. Also, note that transfer learning model performs the best on the pascal dataset. We analyse the performance of these using loss, accuracy and IoU plots along with segmentation maps, which help us draw valuable insights about the working of the models.
The convective heat transfer in a turbulent boundary layer (TBL) on a flat plate is enhanced using an artificial intelligence approach based on linear genetic algorithms control (LGAC). The actuator is a set of six slot jets in crossflow aligned with the freestream. An open-loop optimal periodic forcing is defined by the carrier frequency, the duty cycle and the phase difference between actuators as control parameters. The control laws are optimised with respect to the unperturbed TBL and to the actuation with a steady jet. The cost function includes the wall convective heat transfer rate and the cost of the actuation. The performance of the controller is assessed by infrared thermography and characterised also with particle image velocimetry measurements. The optimal controller yields a slightly asymmetric flow field. The LGAC algorithm converges to the same frequency and duty cycle for all the actuators. It is noted that such frequency is strikingly equal to the inverse of the characteristic travel time of large-scale turbulent structures advected within the near-wall region. The phase difference between multiple jet actuation has shown to be very relevant and the main driver of flow asymmetry. The results pinpoint the potential of machine learning control in unravelling unexplored controllers within the actuation space. Our study furthermore demonstrates the viability of employing sophisticated measurement techniques together with advanced algorithms in an experimental investigation.
The image captioning task is typically realized by an auto-regressive method that decodes the text tokens one by one. We present a diffusion-based captioning model, dubbed the name DDCap, to allow more decoding flexibility. Unlike image generation, where the output is continuous and redundant with a fixed length, texts in image captions are categorical and short with varied lengths. Therefore, naively applying the discrete diffusion model to text decoding does not work well, as shown in our experiments. To address the performance gap, we propose several key techniques including best-first inference, concentrated attention mask, text length prediction, and image-free training. On COCO without additional caption pre-training, it achieves a CIDEr score of 117.8, which is +5.0 higher than the auto-regressive baseline with the same architecture in the controlled setting. It also performs +26.8 higher CIDEr score than the auto-regressive baseline (230.3 v.s.203.5) on a caption infilling task. With 4M vision-language pre-training images and the base-sized model, we reach a CIDEr score of 125.1 on COCO, which is competitive to the best well-developed auto-regressive frameworks. The code is available at https://github.com/buxiangzhiren/DDCap.
Large Vision-Language Foundation Models (VLFM), such as CLIP, ALIGN and Florence, are trained on large-scale datasets of image-caption pairs and achieve superior transferability and robustness on downstream tasks, but they are difficult to use in many practical applications due to their large size, high latency and fixed architectures. Unfortunately, recent work shows training a small custom VLFM for resource-limited applications is currently very difficult using public and smaller-scale data. In this paper, we introduce a new distillation mechanism (DIME-FM) that allows us to transfer the knowledge contained in large VLFMs to smaller, customized foundation models using a relatively small amount of inexpensive, unpaired images and sentences. We transfer the knowledge from the pre-trained CLIP-ViTL/14 model to a ViT-B/32 model, with only 40M public images and 28.4M unpaired public sentences. The resulting model "Distill-ViT-B/32" rivals the CLIP-ViT-B/32 model pre-trained on its private WiT dataset (400M image-text pairs): Distill-ViT-B/32 achieves similar results in terms of zero-shot and linear-probing performance on both ImageNet and the ELEVATER (20 image classification tasks) benchmarks. It also displays comparable robustness when evaluated on five datasets with natural distribution shifts from ImageNet.
We propose a second order gradient based method with ADAM and RMSprop for the training of generative adversarial networks. The proposed method is fastest to obtain similar accuracy when compared to prominent second order methods. Unlike state-of-the-art recent methods, it does not require solving a linear system, or it does not require additional mixed second derivative terms. We derive the fixed point iteration corresponding to proposed method, and show that the proposed method is convergent. The proposed method produces better or comparable inception scores, and comparable quality of images compared to other recently proposed state-of-the-art second order methods. Compared to first order methods such as ADAM, it produces significantly better inception scores. The proposed method is compared and validated on popular datasets such as FFHQ, LSUN, CIFAR10, MNIST, and Fashion MNIST for image generation tasks\footnote{Accepted in IJCNN 2023}. Codes: \url{https://github.com/misterpawan/acom}
Large training data and expensive model tweaking are standard features of deep learning for images. As a result, data owners often utilize cloud resources to develop large-scale complex models, which raises privacy concerns. Existing solutions are either too expensive to be practical or do not sufficiently protect the confidentiality of data and models. In this paper, we study and compare novel \emph{image disguising} mechanisms, DisguisedNets and InstaHide, aiming to achieve a better trade-off among the level of protection for outsourced DNN model training, the expenses, and the utility of data. DisguisedNets are novel combinations of image blocktization, block-level random permutation, and two block-level secure transformations: random multidimensional projection (RMT) and AES pixel-level encryption (AES). InstaHide is an image mixup and random pixel flipping technique \cite{huang20}. We have analyzed and evaluated them under a multi-level threat model. RMT provides a better security guarantee than InstaHide, under the Level-1 adversarial knowledge with well-preserved model quality. In contrast, AES provides a security guarantee under the Level-2 adversarial knowledge, but it may affect model quality more. The unique features of image disguising also help us to protect models from model-targeted attacks. We have done an extensive experimental evaluation to understand how these methods work in different settings for different datasets.