Benchmarking and Improving Detail Image Caption

Code and data for paper:

Benchmarking and Improving Detail Image Caption. Hongyuan Dong*, Jiawen Li*, Bohong Wu, Jiacong Wang, Yuan Zhang, Haoyuan Guo (* Equal Contribution)

Our paper is now available on arXiv.

Overview

Image captioning has long been regarded as a fundamental task in visual understanding. Recently, however, few large vision-language model (LVLM) research discusses model's image captioning performance because of the outdated short-caption benchmarks and unreliable evaluation metrics. In this work, we propose to benchmark detail image caption task by curating high-quality evaluation datasets annotated by human experts, GPT-4V and Gemini-1.5-Pro. We also design a more reliable caption evaluation metric called CAPTURE (CAPtion evaluation by exTracting and coUpling coRE information). CAPTURE extracts visual elements, e.g., objects, attributes and relations from captions, and then matches these elements through three stages, achieving the highest consistency with expert judgements over other rule-based or model-based caption metrics. The proposed benchmark and metric provide reliable evaluation for LVLM's detailed image captioning ability. Guided by this evaluation, we further explore to unleash LVLM's detail caption capabilities by synthesizing high-quality data through a five-stage data construction pipeline. Our pipeline only uses a given LVLM itself and other open-source tools, without any human or GPT-4V annotation in the loop. Experiments show that the proposed data construction strategy significantly improves model-generated detail caption data quality for LVLMs with leading performance, and the data quality can be further improved in a self-looping paradigm.

Detail Image Caption Benchmark

We release the DetailCaps-4870 benchmark, which contains 4870 images with high-quality reference captions annotated by GPT-4V&Gemini-1.5-Pro. The statistics of DetailCaps-4870 compared with other image caption benchmarks of comparables sizes is shown below:

Benchmark	Data source	Annt. expert	Img num	ref num	Avg len	Uni. 2-gram
COCOtest	COCO	Human	$5000$	$25,010$	$10.59$	$61,448$
Nocapsval	Openimages	Human	$4500$	$45,000$	$11.49$	$116,969$
DetailCaps-100	COCO, SAM, LAION, CC, SBU	GPT-4V, Human	$100$	$100$	$175.96$	$10,858$
DetailCaps-4870	COCO, SAM, LAION, CC, SBU, Coyo, Flickr	GPT-4V, GPT4O, Gemini-1.5-Pro	$4870$	$14610$	$122.06$	$533,201$

The evaluation dataset will soon be available on Huggingface. Please download the dataset and put it under the datasets folder.

Detail Image Caption Evaluation Metric: CAPTURE

The proposed metric CAPTURE (CAPtion evaluation by exTracting and coUpling coRE information) achieves the highest consistency with expert judgements on DetailCaps benchmarks. We show the average consistency scores on DetailCaps-100 and DetailCaps-4870 benchmarks in the table below.

Caption metric	PCC $\rho$ $\uparrow$	$1-R^2$ $\downarrow$	Kendall's $\tau$ $\uparrow$	Sample $\tau$ $\uparrow$
BLEU	$0.2608$	$54.75$	$0.1866$	$0.2462$
ROUGE-L	$0.2951$	$134.12$	$0.2149$	$0.3383$
CIDEr	$0.1148$	$2.6e^7$	$0.1165$	$0.0991$
METEOR	$0.4022$	$290.38$	$0.2927$	$0.4062$
SPICE	$0.4386$	$155.95$	$0.3244$	$0.4718$
CLIPScore	$0.3558$	$21.46$	$0.2479$	$0.3841$
CAPTURE	$0.5091$	$8.29$	$0.3861$	$0.6018$

We evaluate SOTA open-source LVLMs' detail captioning abilities with our benchmark and metric. The results are listed below.

Model	Language Model	Caption Data	Resolution	CAPTURE
CogVLM	Vicuna-7B	Human Annt.	$490^2$	$60.06$
ShareCaptioner-7B	Vicuna-7B	GPT-4V Annt.	$448^2$	$59.80$
LLaVA-1.5-7B	Vicuna-7B	Synthesized	$336^2$	$51.05$
LLaVA-1.5-13B	Vicuna-13B	Synthesized	$336^2$	$51.20$
LLaVA-NEXT-7B	Vicuna-7B	GPT-4V Annt.	$336^2$*{1-5}	$58.61$
LLaVA-NEXT-13B	Vicuna-13B	GPT-4V Annt.	$336^2$*{1-5}	$59.01$
LLaVA-NEXT-34B	Hermes-2-Yi-34B	GPT-4V Annt.	$336^2$*{1-5}	$59.20$
Mini-Gemini-HD-7B	Vicuna-7B	GPT-4V Annt.	$336^2$*5	$57.95$
Mini-Gemini-HD-13B	Vicuna-13B	GPT-4V Annt.	$336^2$*5	$58.66$
Intern-XComposerV2	Vicuna-7B	GPT-4V Annt.	$490^2$	$59.86$
InternVL-V1.2-PLUS-40B	Hermes-2-Yi-34B	GPT-4V Annt.	$448^2$	$60.69$
InternVL-V1.5-26B	InternLM-20B	GPT-4V Annt.	$448^2$*{1-41}	$63.42$

Detail Image Caption Construction

We construct a data construction pipeline to unleash LVLM's detail image captioning ability with open-source vision and language tools. We show the performance of the performance of the proposed data construction pipeline with different LVLM bachbones below.

Caption	DetailCaps-100	DetailCaps-4870	Average
LLaVA-1.5-7B self	$51.23$	$51.05$	$51.14$
LLaVA-1.5-7B syn	$57.11$	$56.25$	$56.68$
LLaVA-1.5-13B self	$51.76$	$51.20$	$51.48$
LLaVA-1.5-13B syn	$57.36$	$57.05$	$57.20$
LLaVA-NEXT-7B self	$61.48$	$58.61$	$60.73$
LLaVA-NEXT-7B syn	$62.24$	$60.39$	$61.31$
Mini-Gemini-7B-HD self	$59.51$	$57.95$	$58.73$
Mini-Gemini-7B-HD syn	$60.44$	$59.07$	$59.75$

Quick Start

Environment

Run the following scripts to prepare the environment for CAPTURE and the data construction pipeline.

conda create -n detailcaption python=3.9
conda activate detailcaption
bash prepare.sh

Detail Image Caption Evaluation

We have wrapped the proposed CAPTURE evaluation metric into pip package, and you can install it as follows:

pip3 install capture_metric

After installation, CAPTURE metric can be used in the same way as other caption evaluation metrics implemented in pycocoevalcap, such as BLEU, CIDEr, METEOR, ROUGE, etc. Here is an example:

from capture_metric.capture import CAPTURE
refs = {
  <sample_key>: [ref_0, ref_1, ...],
  ...
}
preds = {
  <sample_key>: [pred_caption],
  ...
}

evaluator = CAPTURE()
score = evaluator.compute_score(refs, preds)
print(f"CAPTURE score: {score}")

You can now use lmms_eval to evaluate you LVLM's detail image caption performance on the DetailCaps-4870 benchmark with CAPTURE metric. We refer to lmms detailcaps for more details.

Detail Image Caption Construction

For detail image caption construction, first download SAM, Owlv2, LLaVA-v1.5 (or other LVLM), LLaMA-2 and place them under ckpt folder:

ckpt
├─sam
|  ├─sam_vit_h_4b8939.pth
|  └─sam_vit_l_0b3195.pth
├─owlv2-large-patch14-ensemble
├─llava-v1.5-13b
├─llava-v1.5-7b
├─llava-v1.5-13b
├─Llama-2-7b-chat-hf
└─Llama-2-13b-chat-hf

Then organize your image data in .parquet format with binary image stored in the frame field. Run the followig script to generate annotations for your parquet data files stored in <source_path>. <model_size> should be set as either 7b or 13b, corresponding to pipelines for different model size.

bash generate_all_annotations.sh <model_size> <source_path>

Citation

@article{dong2024benchmarking,
  title={Benchmarking and Improving Detail Image Caption},
  author={Dong, Hongyuan and Li, Jiawen and Wu, Bohong and Wang, Jiacong and Zhang, Yuan and Guo, Haoyuan},
  journal={arXiv preprint arXiv:2405.19092},
  year={2024}
}