# ShotCrop3: Cropping Human-Centric Images into Cinematic Triple-Shot Compositions

arXiv: https://arxiv.org/abs/2606.05635 · HTML: https://mlsanity.com/paper/2606.05635/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 30 references that are in the map.
- First posted on arXiv 4 Jun 2026.

## Builds on (most cited first, 22 of 22)

- [Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilities](https://mlsanity.com/paper/2507.06261/) (2025)
- [Qwen3-VL Technical Report](https://mlsanity.com/paper/2511.21631/) (2025)
- [InternVL3.5: Advancing Open-Source Multimodal Models in Versatility, Reasoning, and Efficiency](https://mlsanity.com/paper/2508.18265/) (2025)
- [OpenAI GPT-5 System Card](https://mlsanity.com/paper/2601.03267/) (2025)
- [Shikra: Unleashing Multimodal LLM's Referential Dialogue Magic](https://mlsanity.com/paper/2306.15195/) (2023)
- [LISA: Reasoning Segmentation via Large Language Model](https://mlsanity.com/paper/2308.00692/) (2023)
- [YOLO-World: Real-Time Open-Vocabulary Object Detection](https://mlsanity.com/paper/2401.17270/) (2024)
- [Groma: Localized Visual Tokenization for Grounding Multimodal Large Language Models](https://mlsanity.com/paper/2404.13013/) (2024)
- [DINO-X: A Unified Vision Model for Open-World Object Detection and Understanding](https://mlsanity.com/paper/2411.14347/) (2024)
- [Reliable and Efficient Image Cropping: A Grid Anchor based Approach](https://mlsanity.com/paper/1904.04441/) (2019)
- [UniPercept: Towards Unified Perceptual-Level Image Understanding across Aesthetics, Quality, Structure, and Texture](https://mlsanity.com/paper/2512.21675/) (2025)
- [Spatial-Semantic Collaborative Cropping for User Generated Content](https://mlsanity.com/paper/2401.08086/) (2024)
- [Learning Subject-Aware Cropping by Outpainting Professional Photos](https://mlsanity.com/paper/2312.12080/) (2023)
- [Human-centric Image Cropping with Partition-aware and Content-preserving Features](https://mlsanity.com/paper/2207.10269/) (2022)
- [Beyond Image Borders: Learning Feature Extrapolation for Unbounded Image Composition](https://mlsanity.com/paper/2309.12042/) (2023)
- [PhotoFramer: Multi-modal Image Composition Instruction](https://mlsanity.com/paper/2512.00993/) (2025)
- [ProCrop: Learning Aesthetic Image Cropping from Professional Compositions](https://mlsanity.com/paper/2505.22490/) (2025)
- [ClipCrop: Conditioned Cropping Driven by Vision-Language Model](https://mlsanity.com/paper/2211.11492/) (2022)
- [Find Beauty in the Rare: Contrastive Composition Feature Clustering for Nontrivial Cropping Box Regression](https://mlsanity.com/paper/2302.08662/) (2023)
- [Zoomer: Adaptive Image Focus Optimization for Black-box MLLM](https://mlsanity.com/paper/2505.00742/) (2025)
- [Image Cropping under Design Constraints](https://mlsanity.com/paper/2310.08892/) (2023)
- [AesCrop: Aesthetic-driven Cropping Guided by Composition](https://mlsanity.com/paper/2510.22528/) (2025)

## Abstract (arXiv)

Prior work on aesthetic composition typically produces a single aesthetically pleasing crop, overlooking the narrative value of composing multiple shots from one scene. In practice, multi-shot composition is critical for downstream creative workflows: commercial posters often require multiple crops with different emphases (e.g., context, subject, and emotion/product details) to present key story beats. Therefore, we propose Triple-Shot Compositions (TSC), a composition task that generates a three-shot set -- establishing, medium, and close-up -- from a single human-centric image, each paired with a brief shot description to support visual narration. To learn TSC with limited expert annotations, we introduce ShotCrop which undergoes a three-stage training process: it first applies Chain-of-Thought supervised fine-tuning to establish basic reasoning and aesthetic shot-cropping skills, then performs semi-supervised fine-tuning with high-confidence pseudo labels to further enhance aesthetic capability, and is finally optimized with Group Relative Policy Optimization for ShotCrop (GRPO-S) using a composite reward tailored for it. Specifically, our pseudo-labeling strategy combines MLLM-based scoring, aesthetic assessment, and CLIP similarity to retain high-confidence training signals. In addition, we present TSC-Bench, a benchmark of 1.2k expert-annotated test cases. Notably, ShotCrop achieves an average improvement of 2.82 times over GPT-5 in shot localization accuracy.