c-guevara / collavo

Official PyTorch Implementation code for realizing the technical part of CoLLaVO: Crayon Large Language and Vision mOdel to significantly improve zero-shot vision language performances (ACL 2024 Findings)

CoLLaVO: Crayon Large Language and Vision mOdel [arxiv]

📰 News

• CoLLaVO is now available in 🤗Huggingface Space.
• CoLLaVO is featured by Huggingface Daily Papers.
• A new model, MoAI is either released in [Paper]/[Github]/[Huggingface].

🎨 In-Progress

• Code is public (Only Inference Supported).
• Downloading CoLLaVO-7B is available in Huggingface.
• Huggingface README.md for simple running
• Short running code for an image example is available.
• Uploading GPT-Aided Evaluation

---

Official PyTorch implementation code for realizing the technical part of Crayon Large Language and Vision mOdel (CoLLaVO) to improve performance of numerous zero-shot vision language tasks. This code is developed on two baseline codes of XDecoder: Generalized Decoding for Pixel, Image, and Language accepted in CVPR 2023 and InternLM for Technical Paper.

🏝️ Summary

The remarkable success of Large Language Models (LLMs) and instruction tuning drives the evolution of Vision Language Models (VLMs) towards a versatile general-purpose model. Yet, it remains unexplored whether current VLMs genuinely possess quality object-level image understanding capabilities determined from 'what objects are in the image?' or 'which object corresponds to a specified bounding box?'. Our findings reveal that the image understanding capabilities of current VLMs are strongly correlated with their zero-shot performance on vision language (VL) tasks. This suggests that prioritizing basic image understanding is crucial for VLMs to excel at VL tasks. To enhance object-level image understanding, we propose Crayon Large Language and Vision mOdel ( CoLLaVO), which incorporates instruction tuning with Crayon Prompt as a new visual prompt tuning scheme based on panoptic color maps. Furthermore, we present a learning strategy of Dual QLoRA to preserve object-level image understanding without forgetting it during visual instruction tuning, thereby achieving a significant leap in numerous VL benchmarks in a zero-shot setting.

🚀 Highlights

Figure. Zero-shot performance of CoLLaVO-7B on challenging VL datasets compared with closed-source VLMs: GPT-4V, Gemini-Pro, Qwen-VL-Plus. Note: The scores of MME are rescaled by 1/20 to match the scales with the accuracies of others.

Figure. Demonstrating the efficiency and effectiveness of CoLLaVO compared with those of other VLMs. Note that accuracy is measured on SEED-IMG.

Table. Measuring four metrics: Accuracy, Precision, Recall, F1-score on three types of question answering to evaluate hallucination of vision language models: Adversarial, Random, and Popular in POPE.

📖 Citation

@article{lee2024collavo,
  title={CoLLaVO: Crayon Large Language and Vision mOdel},
  author={Lee, Byung-Kwan and Park, Beomchan and Kim, Chae Won and Ro, Yong Man},
  journal={arXiv preprint arXiv:2402.11248},
  year={2024}
}

Download CoLLaVO-7B (Under Preparation)

	GQA	SQA-IMG	TextVQA	POPE	MME-P	MME-C	MM-Bench	MMB-CN	MM-Vet	Q-Bench
BLIP2-13B	42.4	61.0	42.5	85.3	1293.8	290.0	-	-	22.4	-
InstructBLIP-7B	49.5	49.2	60.5	50.1	-	-	36.0	23.7	25.6	56.7
Qwen-VL-Chat-7B	57.5	68.2	61.5	-	1487.5	360.7	60.6	56.7	-	-
LLaVA1.5-7B	62.0	66.8	58.2	85.9	1510.7	293.8	64.3	58.3	30.5	58.7
CoLLaVO-7B	61.4	80.7	64.2	87.2	1689.7	525.0	83.0	82.1	40.3	67.6

📂 Directory Layout

.
├── asset                           # Required package lists (Important)
├── trainer                         # Training CoLLaVO and initializing optimizer (Not Support Now)
├── utils                           # Michallengeous util files (Not important)
├── collavo                         # CoLLaVO architecture & loading collavo (Important)
├── pipeline                        # Evaluating zero-shot vision language tasks (Important)
│
├── datasets                        # Important
│   ├── dataset_mappers             # data parsing including augmentation for loader
│   ├── evaluation                  # measuring evaluation for each dataset 
│   └── registration                # register dataset
│
├── configs                         
│   ├── accel                       # Accelerate Config files (Support DDP)
│   └── collavo_eval.yaml           # Config of evaluating collavo
│
├── modeling                        # Not Important
│   ├── architectures               # training the prototype of collavo (Not Support Now)
│   ├── utils                       # utils for modeling (Not important)
│   └── BaseModel                   # loading and saving model
│
├── lbk_entry.py                    # main code of control tower (Important)
├── run                             # bash file for running the evaluation (Important)
│
├── install                         # install required packages (Important)
└── README.md

---

💡 How to Run?

In bash file of install, you should first run the following lines.

conda create -n collavo python=3.9
conda activate collavo
conda clean -a && pip cache purge
conda install pytorch==2.0.1 torchvision==0.15.2 torchaudio==2.0.2 pytorch-cuda=11.8 -c pytorch -c nvidia
pip install -r assets/requirements/requirements.txt
pip install -r assets/requirements/requirements_custom.txt
pip install flash-attn --no-build-isolation

In addition, you should set the following environment variables to set the dataset path.

export DETECTRON2_DATASETS=/path/to/dataset
export DATASET=/path/to/dataset
export DATASET2=/path/to/dataset
export VLDATASET=/path/to/dataset

Download CoLLaVO-7B Model and then you can run the demo.py

"""
CoLLaVO-7B

Simple Six Steps
"""

# [1] Loading Image
from PIL import Image
from torchvision.transforms import Resize
from torchvision.transforms.functional import pil_to_tensor
image_path = "figures/crayon_image.jpg"
image = Resize(size=(490, 490), antialias=False)(pil_to_tensor(Image.open(image_path)))

# [2] Instruction Prompt
prompt = "Describe this image in detail"

# [3] Loading CoLLaVO
from collavo.load_collavo import prepare_collavo
collavo_model, collavo_processor, seg_model, seg_processor = prepare_collavo(collavo_path='BK-Lee/CoLLaVO-7B', bits=4, dtype='fp16')

# [4] Pre-processing for CoLLaVO
collavo_inputs = collavo_model.demo_process(image=image, 
                                    prompt=prompt, 
                                    processor=collavo_processor,
                                    seg_model=seg_model,
                                    seg_processor=seg_processor,
                                    device='cuda:0')

# [5] Generate
import torch
with torch.inference_mode():
    generate_ids = collavo_model.generate(**collavo_inputs, do_sample=True, temperature=0.9, top_p=0.95, max_new_tokens=256, use_cache=True)

# [6] Decoding
answer = collavo_processor.batch_decode(generate_ids, skip_special_tokens=True)[0].split('[U')[0]
print(answer)

If you want to valiate zero-shot performances in numerous datasets, then running the bash file 'run'.

# CoLLaVO-Experiment
GPU_DEVICE="0,1,2,3,4,5"
length=${#GPU_DEVICE}
n_gpu=$(((length+1)/2))
main_port=10000
test_batch=1

CUDA_VISIBLE_DEVICES=$GPU_DEVICE \
accelerate launch --config_file configs/accel/ddp_accel.yaml \
    --num_processes=$n_gpu \
    --main_process_port=$main_port \
    lbk_entry.py eval \
    --conf_files configs/collavo_eval.yaml \
    --overrides \
    WANDB False \
    DATASETS.TEST mme \
    PIPELINE MMEPipeline \
    MME.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    SCIENCEQA.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    POPE.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    MMBENCH.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    MMVET.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    AI2D.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    HALLUSIONBENCH.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    MATHVISTA.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    QBENCH.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    SEED.TEST.BATCH_SIZE_TOTAL $((n_gpu * test_batch)) \
    SAVE_DIR /path/to/CoLLaVO_DIR \
    WEIGHT True \
    RESUME_FROM /path/to/CoLLaVO_WEIGHT \

Note that, you should change the two parts to evaluate the dataset you want. (This is very important!!)

DATASETS.TEST

• GQA: gqa_testdev_balanced
• SQA-IMG: scienceqa_test
• TextVQA: textvqa_val
• POPE: pope_test
• MME: mme
• MM-Bench: mmbench or mmbench_cn
• MM-Vet: mm-vet
• Q-Bench: qbench_dev
• MATHVISTA: mathvista_testmini
• AI2D: ai2d
• SEED-IMG: seed
• HallusionBench: hallusionbench

PIPELINE

• GQA: GQAPipeline
• SQA-IMG: SQAPipeline
• TextVQA: TextVQAPipeline
• POPE: POPEPipeline
• MME: MMEPipeline
• MM-Bench: MMBenchPipeline
• MM-Vet: MMVetPipeline
• Q-Bench: QBenchPipeline
• MATHVISTA: MathVistaPipeline
• AI2D: AI2DPipeline
• SEED-IMG: SEEDPipeline
• HallusionBench: HallusionPipeline

GPT-4 Aid Evalution for AI2D, MM-Vet, SEED-IMG

This code will be soon public!

🍅 Download Datasets

• GQA
• SQA-IMG
• TextVQA
• POPE
• MME
• MM-Bench
• MM-Vet
• Q-Bench
• MathVista
• AI2D
• SEED-IMG
• HallusionBench

📂 Dataset Directory (/path/to/dataset)

.
├── GQA                             # GQA
├── ScienceQA                       # SQA-IMG
├── TextVQA                         # TextVQA
├── POPE                            # POPE
├── MME_Benchmark_release_version   # MME
├── MMBench                         # MM-Bench
├── mm-vet                          # MM-Vet
├── LLVisionQA-QBench               # Q-Bench
├── MathVista                       # MathVista
├── SEED-Bench                      # SEED-IMG
├── ai2d                            # AI2D
└── HallusionBench                  # HallusionBench