File size: 16,315 Bytes
33ee8e1 15aa04e 33ee8e1 00d2f15 33ee8e1 6bf1792 33ee8e1 6bf1792 33ee8e1 6bf1792 33ee8e1 6bf1792 33ee8e1 00d2f15 33ee8e1 6bf1792 33ee8e1 6bf1792 33ee8e1 6bf1792 33ee8e1 0f7c938 6bf1792 33ee8e1 0f7c938 33ee8e1 0f7c938 33ee8e1 0f7c938 33ee8e1 0f7c938 33ee8e1 6e0aa88 33ee8e1 0f7c938 33ee8e1 0f7c938 33ee8e1 0f7c938 33ee8e1 6e0aa88 0c7be59 6e0aa88 0c7be59 f0acedb 33ee8e1 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 |
---
license: mit
license_link: https://huggingface.co/microsoft/Florence-2-large/resolve/main/LICENSE
pipeline_tag: image-text-to-text
tags:
- vision
---
# Florence-2: Advancing a Unified Representation for a Variety of Vision Tasks
## Model Summary
**This is a continued pretrained version of Florence-2-large model with 4k context length, only 0.1B samples are used for continue pretraining, thus it might not be trained well. In addition, OCR task has been updated with line separator ('\n'). COCO OD AP 39.8**
This Hub repository contains a HuggingFace's `transformers` implementation of Florence-2 model from Microsoft.
Florence-2 is an advanced vision foundation model that uses a prompt-based approach to handle a wide range of vision and vision-language tasks. Florence-2 can interpret simple text prompts to perform tasks like captioning, object detection, and segmentation. It leverages our FLD-5B dataset, containing 5.4 billion annotations across 126 million images, to master multi-task learning. The model's sequence-to-sequence architecture enables it to excel in both zero-shot and fine-tuned settings, proving to be a competitive vision foundation model.
Resources and Technical Documentation:
+ [Florence-2 technical report](https://arxiv.org/abs/2311.06242).
+ [Jupyter Notebook for inference and visualization of Florence-2-large](https://huggingface.co/microsoft/Florence-2-large/blob/main/sample_inference.ipynb)
| Model | Model size | Model Description |
| ------- | ------------- | ------------- |
| Florence-2-base[[HF]](https://huggingface.co/microsoft/Florence-2-base) | 0.23B | Pretrained model with FLD-5B
| Florence-2-large[[HF]](https://huggingface.co/microsoft/Florence-2-large) | 0.77B | Pretrained model with FLD-5B
| Florence-2-base-ft[[HF]](https://huggingface.co/microsoft/Florence-2-base-ft) | 0.23B | Finetuned model on a colletion of downstream tasks
| Florence-2-large-ft[[HF]](https://huggingface.co/microsoft/Florence-2-large-ft) | 0.77B | Finetuned model on a colletion of downstream tasks
## How to Get Started with the Model
Use the code below to get started with the model. All models are trained with float16.
```python
import requests
import torch
from PIL import Image
from transformers import AutoProcessor, AutoModelForCausalLM
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
model = AutoModelForCausalLM.from_pretrained("microsoft/Florence-2-large", torch_dtype=torch_dtype, trust_remote_code=True).to(device)
processor = AutoProcessor.from_pretrained("microsoft/Florence-2-large", trust_remote_code=True)
prompt = "<OD>"
url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/car.jpg?download=true"
image = Image.open(requests.get(url, stream=True).raw)
inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)
generated_ids = model.generate(
input_ids=inputs["input_ids"],
pixel_values=inputs["pixel_values"],
max_new_tokens=4096,
num_beams=3,
do_sample=False
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
parsed_answer = processor.post_process_generation(generated_text, task="<OD>", image_size=(image.width, image.height))
print(parsed_answer)
```
## Tasks
This model is capable of performing different tasks through changing the prompts.
First, let's define a function to run a prompt.
<details>
<summary> Click to expand </summary>
```python
import requests
import torch
from PIL import Image
from transformers import AutoProcessor, AutoModelForCausalLM
device = "cuda:0" if torch.cuda.is_available() else "cpu"
torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
model = AutoModelForCausalLM.from_pretrained("microsoft/Florence-2-large", torch_dtype=torch_dtype, trust_remote_code=True).to(device)
processor = AutoProcessor.from_pretrained("microsoft/Florence-2-large", trust_remote_code=True)
url = "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/transformers/tasks/car.jpg?download=true"
image = Image.open(requests.get(url, stream=True).raw)
def run_example(task_prompt, text_input=None):
if text_input is None:
prompt = task_prompt
else:
prompt = task_prompt + text_input
inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)
generated_ids = model.generate(
input_ids=inputs["input_ids"],
pixel_values=inputs["pixel_values"],
max_new_tokens=1024,
num_beams=3
)
generated_text = processor.batch_decode(generated_ids, skip_special_tokens=False)[0]
parsed_answer = processor.post_process_generation(generated_text, task=task_prompt, image_size=(image.width, image.height))
print(parsed_answer)
```
</details>
Here are the tasks `Florence-2` could perform:
<details>
<summary> Click to expand </summary>
### Caption
```python
prompt = "<CAPTION>"
run_example(prompt)
```
### Detailed Caption
```python
prompt = "<DETAILED_CAPTION>"
run_example(prompt)
```
### More Detailed Caption
```python
prompt = "<MORE_DETAILED_CAPTION>"
run_example(prompt)
```
### Caption to Phrase Grounding
caption to phrase grounding task requires additional text input, i.e. caption.
Caption to phrase grounding results format:
{'\<CAPTION_TO_PHRASE_GROUNDING>': {'bboxes': [[x1, y1, x2, y2], ...], 'labels': ['', '', ...]}}
```python
task_prompt = "<CAPTION_TO_PHRASE_GROUNDING>"
results = run_example(task_prompt, text_input="A green car parked in front of a yellow building.")
```
### Object Detection
OD results format:
{'\<OD>': {'bboxes': [[x1, y1, x2, y2], ...],
'labels': ['label1', 'label2', ...]} }
```python
prompt = "<OD>"
run_example(prompt)
```
### Dense Region Caption
Dense region caption results format:
{'\<DENSE_REGION_CAPTION>' : {'bboxes': [[x1, y1, x2, y2], ...],
'labels': ['label1', 'label2', ...]} }
```python
prompt = "<DENSE_REGION_CAPTION>"
run_example(prompt)
```
### Region proposal
Dense region caption results format:
{'\<REGION_PROPOSAL>': {'bboxes': [[x1, y1, x2, y2], ...],
'labels': ['', '', ...]}}
```python
prompt = "<REGION_PROPOSAL>"
run_example(prompt)
```
### OCR
```python
prompt = "<OCR>"
run_example(prompt)
```
### OCR with Region
OCR with region output format:
{'\<OCR_WITH_REGION>': {'quad_boxes': [[x1, y1, x2, y2, x3, y3, x4, y4], ...], 'labels': ['text1', ...]}}
```python
prompt = "<OCR_WITH_REGION>"
run_example(prompt)
```
### Output confidence score with Object Detection
```python
def run_example_with_score(task_prompt, text_input=None):
if text_input is None:
prompt = task_prompt
else:
prompt = task_prompt + text_input
inputs = processor(text=prompt, images=image, return_tensors="pt").to(device, torch_dtype)
generated_ids = model.generate(
input_ids=inputs["input_ids"],
pixel_values=inputs["pixel_values"],
max_new_tokens=1024,
num_beams=3,
return_dict_in_generate=True,
output_scores=True,
)
generated_text = processor.batch_decode(generated_ids.sequences, skip_special_tokens=False)[0]
prediction, scores, beam_indices = generated_ids.sequences, generated_ids.scores, generated_ids.beam_indices
transition_beam_scores = model.compute_transition_scores(
sequences=prediction,
scores=scores,
beam_indices=beam_indices,
)
parsed_answer = processor.post_process_generation(sequence=generated_ids.sequences[0],
transition_beam_score=transition_beam_scores[0],
task=task_prompt, image_size=(image.width, image.height)
)
print(parsed_answer)
prompt = "<OD>"
run_example_with_score(prompt)
```
for More detailed examples, please refer to [notebook](https://huggingface.co/microsoft/Florence-2-large/blob/main/sample_inference.ipynb)
</details>
# Benchmarks
## Florence-2 Zero-shot performance
The following table presents the zero-shot performance of generalist vision foundation models on image captioning and object detection evaluation tasks. These models have not been exposed to the training data of the evaluation tasks during their training phase.
| Method | #params | COCO Cap. test CIDEr | NoCaps val CIDEr | TextCaps val CIDEr | COCO Det. val2017 mAP |
|--------|---------|----------------------|------------------|--------------------|-----------------------|
| Flamingo | 80B | 84.3 | - | - | - |
| Florence-2-base| 0.23B | 133.0 | 118.7 | 70.1 | 34.7 |
| Florence-2-large| 0.77B | 135.6 | 120.8 | 72.8 | 37.5 |
The following table continues the comparison with performance on other vision-language evaluation tasks.
| Method | Flickr30k test R@1 | Refcoco val Accuracy | Refcoco test-A Accuracy | Refcoco test-B Accuracy | Refcoco+ val Accuracy | Refcoco+ test-A Accuracy | Refcoco+ test-B Accuracy | Refcocog val Accuracy | Refcocog test Accuracy | Refcoco RES val mIoU |
|--------|----------------------|----------------------|-------------------------|-------------------------|-----------------------|--------------------------|--------------------------|-----------------------|------------------------|----------------------|
| Kosmos-2 | 78.7 | 52.3 | 57.4 | 47.3 | 45.5 | 50.7 | 42.2 | 60.6 | 61.7 | - |
| Florence-2-base | 83.6 | 53.9 | 58.4 | 49.7 | 51.5 | 56.4 | 47.9 | 66.3 | 65.1 | 34.6 |
| Florence-2-large | 84.4 | 56.3 | 61.6 | 51.4 | 53.6 | 57.9 | 49.9 | 68.0 | 67.0 | 35.8 |
## Florence-2 finetuned performance
We finetune Florence-2 models with a collection of downstream tasks, resulting two generalist models *Florence-2-base-ft* and *Florence-2-large-ft* that can conduct a wide range of downstream tasks.
The table below compares the performance of specialist and generalist models on various captioning and Visual Question Answering (VQA) tasks. Specialist models are fine-tuned specifically for each task, whereas generalist models are fine-tuned in a task-agnostic manner across all tasks. The symbol "▲" indicates the usage of external OCR as input.
| Method | # Params | COCO Caption Karpathy test CIDEr | NoCaps val CIDEr | TextCaps val CIDEr | VQAv2 test-dev Acc | TextVQA test-dev Acc | VizWiz VQA test-dev Acc |
|----------------|----------|-----------------------------------|------------------|--------------------|--------------------|----------------------|-------------------------|
| **Specialist Models** | | | | | | | |
| CoCa | 2.1B | 143.6 | 122.4 | - | 82.3 | - | - |
| BLIP-2 | 7.8B | 144.5 | 121.6 | - | 82.2 | - | - |
| GIT2 | 5.1B | 145.0 | 126.9 | 148.6 | 81.7 | 67.3 | 71.0 |
| Flamingo | 80B | 138.1 | - | - | 82.0 | 54.1 | 65.7 |
| PaLI | 17B | 149.1 | 127.0 | 160.0▲ | 84.3 | 58.8 / 73.1▲ | 71.6 / 74.4▲ |
| PaLI-X | 55B | 149.2 | 126.3 | 147.0 / 163.7▲ | 86.0 | 71.4 / 80.8▲ | 70.9 / 74.6▲ |
| **Generalist Models** | | | | | | | |
| Unified-IO | 2.9B | - | 100.0 | - | 77.9 | - | 57.4 |
| Florence-2-base-ft | 0.23B | 140.0 | 116.7 | 143.9 | 79.7 | 63.6 | 63.6 |
| Florence-2-large-ft | 0.77B | 143.3 | 124.9 | 151.1 | 81.7 | 73.5 | 72.6 |
| Method | # Params | COCO Det. val2017 mAP | Flickr30k test R@1 | RefCOCO val Accuracy | RefCOCO test-A Accuracy | RefCOCO test-B Accuracy | RefCOCO+ val Accuracy | RefCOCO+ test-A Accuracy | RefCOCO+ test-B Accuracy | RefCOCOg val Accuracy | RefCOCOg test Accuracy | RefCOCO RES val mIoU |
|----------------------|----------|-----------------------|--------------------|----------------------|-------------------------|-------------------------|------------------------|---------------------------|---------------------------|------------------------|-----------------------|------------------------|
| **Specialist Models** | | | | | | | | | | | | |
| SeqTR | - | - | - | 83.7 | 86.5 | 81.2 | 71.5 | 76.3 | 64.9 | 74.9 | 74.2 | - |
| PolyFormer | - | - | - | 90.4 | 92.9 | 87.2 | 85.0 | 89.8 | 78.0 | 85.8 | 85.9 | 76.9 |
| UNINEXT | 0.74B | 60.6 | - | 92.6 | 94.3 | 91.5 | 85.2 | 89.6 | 79.8 | 88.7 | 89.4 | - |
| Ferret | 13B | - | - | 89.5 | 92.4 | 84.4 | 82.8 | 88.1 | 75.2 | 85.8 | 86.3 | - |
| **Generalist Models** | | | | | | | | | | | | |
| UniTAB | - | - | - | 88.6 | 91.1 | 83.8 | 81.0 | 85.4 | 71.6 | 84.6 | 84.7 | - |
| Florence-2-base-ft | 0.23B | 41.4 | 84.0 | 92.6 | 94.8 | 91.5 | 86.8 | 91.7 | 82.2 | 89.8 | 82.2 | 78.0 |
| Florence-2-large-ft| 0.77B | 43.4 | 85.2 | 93.4 | 95.3 | 92.0 | 88.3 | 92.9 | 83.6 | 91.2 | 91.7 | 80.5 |
## BibTex and citation info
```
@article{xiao2023florence,
title={Florence-2: Advancing a unified representation for a variety of vision tasks},
author={Xiao, Bin and Wu, Haiping and Xu, Weijian and Dai, Xiyang and Hu, Houdong and Lu, Yumao and Zeng, Michael and Liu, Ce and Yuan, Lu},
journal={arXiv preprint arXiv:2311.06242},
year={2023}
}
``` |