QSCP: Beyond Class-Name Prompts for Query-Guided Semantic Change Parsing

Yuan Qian; Jie Ma

cs.CVSep 27, 2026

QSCP: Beyond Class-Name Prompts for Query-Guided Semantic Change Parsing

Authors: Yuan Qian, Jie Ma

Organizations: School of Information Science and Technology, Beijing Foreign Studies University, Beijing 100089, China

Abstract

Traditional change detection (CD) identifies changes between bi-temporal remote sensing images, while semantic change detection (SCD) assigns predefined land-cover classes. However, mapping all changes may not meet a user's specific needs. Referring change detection (RCD) enables selective retrieval through category prompts. However, existing category-prompted RCD uses the queried category to specify the destination of a change and returns only a binary mask of the corresponding regions. Users may instead request a particular transition and paired semantic maps to understand what changed into what. Such requests require explicit source and target reasoning beyond target-class localization. To address these needs, we propose query-guided semantic change parsing (QSCP), which supports category names, synonyms, and intent-bearing sentences and returns a query-specific mask with paired temporal semantic maps. QSCP parses requests into intents and semantic slots, composes bidirectional visual evidence, and predicts both temporal states with a query-conditioned decoder. On SECOND, QSCP outperforms RCDNet on synonym, sentence, and transition queries and improves end-to-end semantic prediction over evaluated semantic baselines. WHU-CDC experiments further assess cross-dataset transfer and consistency across equivalent expressions without target-domain training. Code is available at https://github.com/qianyuancs/QSCP

Figures & tables

Fig. 1: QSCP overview. Parsed requests control directional evidence composition and temporal semantic decoding. The predicted mask gates both semantic maps.

Method	All-class	Present-class	Synonym	Sentence	Transition	Unknown	Sem.GT	E2E-Sem
RCDNet	71.15	70.56	53.34	53.56	48.56	49.65	N/A	N/A
SCanNet	64.32	64.43	44.87	46.09	48.16	100.00	33.36	22.19
Seg2Change	66.64	65.27	45.10	46.29	48.16	100.00	44.71	23.12
DynamicEarth	59.70	58.16	45.22	46.43	48.16	100.00	N/A	N/A
QSCP (ours)	71.35	70.71	70.71	71.74	70.99	100.00	62.35	38.01

TABLE I: SECOND query comparison (%): suite BIoU and Present-class temporal semantic mIoU. Sem.GT uses GT regions; E2E-Sem includes predicted-mask errors. Bold: best; N/A: unavailable.

Method	Changed IoU	Unchanged IoU	BIoU
SCanNet [ 7 ]	57.47	87.65	72.56
Seg2Change [ 12 ]	48.61	86.84	67.72
DynamicEarth [ 11 ]	38.36	80.83	59.60
Native RCDNet [ 8 ]	57.17	87.60	72.39
QSCP	57.25	87.78	72.51

TABLE II: SECOND whole-image CD (%): one merged map per image. Bold/underline: best/second best.

Method	Canonical	Synonym	Sentence	Mean	Unknown
RCDNet	71.06	62.39	50.07	61.17	54.77
DynamicEarth	60.74	57.96	57.39	58.70	27.92
Seg2Change	78.11	73.32	54.88	68.77	98.92
QSCP	70.94	70.94	70.94	70.94	100.00

TABLE III: WHU-CDC transfer (%). Baselines encode raw text; QSCP uses normalized routes. Mean: supported-suite BIoU; Unknown: rejection accuracy.

Fig. 2: Qualitative comparisons on SECOND. From top to bottom, the queries are “playground-related changes” (Sentence), “paved ground became houses” (Transition), “water surface” (Synonym), “building” (Canonical), and “oil tanks” (Unknown). White indicates the queried change regions. The unknown query tests ontology rejection. Examples are selected for illustration.

Fig. 3: SECOND temporal semantics at T1/T2. Colors denote classes; black denotes background or query-excluded regions.

Variant	Sem. GT	Pair GT	Sem. Int.	Mask Pos.
Full semantic head	71.04	77.66	76.92	47.95
w/o explicit language	66.17	73.20	73.42	47.95
w/o direction features	59.94	65.33	67.46	47.95

TABLE IV: Semantic-head ablation on SECOND validation (%). Int.: prediction–GT intersection; Pair: temporal-pair accuracy. The localizer is frozen.

Semantic Change Detection

Semantic change detection identifies how meanings, concepts, or scene contents differ across time in text, images, and other data.

See all 63 papers

RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-N Ranking
Apr 22, 2026Roie Kazoom, Yotam Gigi, George Leifman +2Multimodal QA Remote Sensing Benchmarks
Google Research
JL1-CC&QA: Extending the JL1-CD Benchmark with Change Captioning and Question Answering
Jun 30, 2026Ziyuan Liu, Ruifei Zhu, Ouqiao Ma +1Remote Sensing Image Understanding Remote Sensing VQA
Department of Electronic Engineering, Beijing National Research Center for Information Science and Technology, Tsinghua University, Beijing 100084, China · Chang Guang Satellite Technology Co., Ltd. (CGSTL) Changchun 130102, China · College of Communications Engineering, Army Engineering University of PLA, Nanjing 210007, China
Selective Tool Use for Agentic Change Visual Question Answering in Remote Sensing
Sep 13, 2026Yakoub Bazi, Mohamad M. Al Rahhal, Mohamed A. Mekhtiche +1Remote Sensing VQA Tool Use in VLMs
Computer Engineering Department, College of Computer and Information Sciences, King Saud University, Riyadh 11543, Saudi Arabia · Applied Computer Science Department, College of Applied Computer Science, King Saud University, Riyadh 11543, Saudi Arabia

Apr 22, 2026J/K move · Enter open · S save

RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-N Ranking

Roie Kazoom, Yotam Gigi, George Leifman, Tomer Shekel+1

Google Research

Multimodal QA Remote Sensing Benchmarks

Open paper AskPDF arXiv

Traditional change detection identifies where changes occur, but does not explain what changed in natural language. Existing remote sensing change captioning datasets typically describe overall image-level differences, leaving fine-grained localized semantic reasoning largely unexplored. To close this gap, we present RSRCC, a new benchmark for remote sensing change question-answering containing 126k questions, split into 87k training, 17.1k validation, and 22k test instances. Unlike prior datasets, RSRCC is built around localized, change-specific questions that require reasoning about a particular semantic change. To the best of our knowledge, this is the first remote sensing change question-answering benchmark designed explicitly for such fine-grained reasoning-based supervision. To construct RSRCC, we introduce a hierarchical semi-supervised curation pipeline that uses Best-of-N ranking as a critical final ambiguity-resolution stage. First, candidate change regions are extracted from semantic segmentation masks, then initially screened using an image-text embedding model, and finally validated through retrieval-augmented vision-language curation with Best-of-N ranking. This process enables scalable filtering of noisy and ambiguous candidates while preserving semantically meaningful changes. The dataset is available at https://huggingface.co/datasets/google/RSRCC.

Abstract

Semantic Change Detection

Explore similar work

RSRCC: A Remote Sensing Regional Change Comprehension Benchmark Constructed via Retrieval-Augmented Best-of-N Ranking

JL1-CC&QA: Extending the JL1-CD Benchmark with Change Captioning and Question Answering

Selective Tool Use for Agentic Change Visual Question Answering in Remote Sensing