TIPS: Text-Induced Pose Synthesis

TL;DR

This work tackles the limitation of keypoint guided pose transfer that requires a target image by introducing a text guided three stage pipeline that converts a textual target pose description into a target keypoint representation, refines facial keypoints, and renders a pose transferred image from a source image using attention guided synthesis. The approach comprises a text to keypoints generator $G_T$ with a WGAN-GP discriminator, a facial keypoints refinement network, and a pose rendering network $G_S$ that fuses $I_A$ with pose heatmaps to produce $\tilde{I}_B$, trained with a combination of $L_{l_1}$, $L_{GAN}$, and perceptual losses at multiple layers. A new dataset DF-PASS is introduced with 40488 DeepFashion images and human pose descriptions, enabling training without access to the actual target image; extensive ablations and a gender consistency metric (GCR) demonstrate competitive performance against traditional baselines and improved generalization to real-world target poses, highlighting the method's potential for practical deployment when the target image is unavailable. Overall, the paper advances text guided pose synthesis by showing that textual descriptions can effectively drive pose transfer with preserved appearance and background, and provides a public dataset and code to foster further research.

Abstract

In computer vision, human pose synthesis and transfer deal with probabilistic image generation of a person in a previously unseen pose from an already available observation of that person. Though researchers have recently proposed several methods to achieve this task, most of these techniques derive the target pose directly from the desired target image on a specific dataset, making the underlying process challenging to apply in real-world scenarios as the generation of the target image is the actual aim. In this paper, we first present the shortcomings of current pose transfer algorithms and then propose a novel text-based pose transfer technique to address those issues. We divide the problem into three independent stages: (a) text to pose representation, (b) pose refinement, and (c) pose rendering. To the best of our knowledge, this is one of the first attempts to develop a text-based pose transfer framework where we also introduce a new dataset DF-PASS, by adding descriptive pose annotations for the images of the DeepFashion dataset. The proposed method generates promising results with significant qualitative and quantitative scores in our experiments.

Figures (13)

• Figure 1: Overview of the proposed approach. Keypoint-guided methods tend to produce structurally inconsistent images when the physical appearance of the target pose reference significantly differs from the condition image. The proposed text-guided technique successfully addresses this issue while retaining the ability to generate visually decent results close to the keypoint-guided baseline.
• Figure 2: Architecture of the proposed pipeline. The workflow is divided into three stages. In stage 1, we estimate a spatial representation $K^*_B$ for the target pose $P_B$ from the corresponding text description embedding $v_B$. In stage 2, we regressively refine the initial estimation of the facial keypoints to obtain the refined target keypoints $\tilde{K}^*_B$. Finally, in stage 3, we render the target image $\tilde{I}_B$ by conditioning the pose transfer on the source image $I_A$ having the keypoints $K_A$ corresponding to the source pose $P_A$.
• Figure 3: Qualitative results of text to pose generation using $G_T$.
• Figure 4: Qualitative results of regressive refinement using $N_R$.
• Figure 5: Qualitative results of different pose transfer algorithms.