Image-Text-to-Text
Transformers
Safetensors
English
qwen3_5
text-generation-inference
VisionGuardrail
JSON
image-classification
content-safety
content-moderation
multimodal
safety-classifier
guardrail
visual-safety
multimodal-content-filter
mmcf
v1.0
conversational
Instructions to use prithivMLmods/VisionGuardrail-4B with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use prithivMLmods/VisionGuardrail-4B with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("image-text-to-text", model="prithivMLmods/VisionGuardrail-4B") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] pipe(text=messages)# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("prithivMLmods/VisionGuardrail-4B") model = AutoModelForMultimodalLM.from_pretrained("prithivMLmods/VisionGuardrail-4B", device_map="auto") messages = [ { "role": "user", "content": [ {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"}, {"type": "text", "text": "What animal is on the candy?"} ] }, ] inputs = processor.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=256) print(processor.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use prithivMLmods/VisionGuardrail-4B with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "prithivMLmods/VisionGuardrail-4B" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "prithivMLmods/VisionGuardrail-4B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker
docker model run hf.co/prithivMLmods/VisionGuardrail-4B
- SGLang
How to use prithivMLmods/VisionGuardrail-4B with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "prithivMLmods/VisionGuardrail-4B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "prithivMLmods/VisionGuardrail-4B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "prithivMLmods/VisionGuardrail-4B" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "prithivMLmods/VisionGuardrail-4B", "messages": [ { "role": "user", "content": [ { "type": "text", "text": "Describe this image in one sentence." }, { "type": "image_url", "image_url": { "url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg" } } ] } ] }' - Docker Model Runner
How to use prithivMLmods/VisionGuardrail-4B with Docker Model Runner:
docker model run hf.co/prithivMLmods/VisionGuardrail-4B
Update README.md
Browse files
README.md
CHANGED
|
@@ -7,7 +7,6 @@
|
|
| 7 |
|
| 8 |
.vg-code, .vg-code * {
|
| 9 |
font-family: 'Ubuntu Mono', 'Courier New', monospace !important;
|
| 10 |
-
line-height: 1.5;
|
| 11 |
}
|
| 12 |
</style>
|
| 13 |
|
|
@@ -25,26 +24,26 @@
|
|
| 25 |
|
| 26 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Key Highlights</h2>
|
| 27 |
|
| 28 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 29 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 30 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 31 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 32 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 33 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 34 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 35 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 36 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 37 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 38 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 39 |
|
| 40 |
<details style="border: 1px solid #333333; border-radius: 4px; margin-bottom: 16px; overflow: hidden; background-color: #000000;">
|
| 41 |
-
<summary style="padding:
|
| 42 |
-
<div style="background-color: #000000; border: none;
|
| 43 |
-
<div style="background-color: #
|
| 44 |
-
<span>text</span>
|
| 45 |
-
<button onclick="copyCode(this)" style="background-color: #
|
| 46 |
</div>
|
| 47 |
-
<pre
|
| 48 |
|
| 49 |
Your role is minor protection: any media containing suggestive, revealing, sexually provocative, or explicit visual elements must be flagged as NSFW.
|
| 50 |
|
|
@@ -81,20 +80,20 @@ Output strictly valid JSON with no extra conversational text, commentary, or mar
|
|
| 81 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Quick Start with Transformers</h2>
|
| 82 |
|
| 83 |
<div style="background-color: #000000; border: 1px solid #333333; border-radius: 4px; margin-bottom: 16px; overflow: hidden;">
|
| 84 |
-
<div style="background-color: #
|
| 85 |
-
<span>bash</span>
|
| 86 |
-
<button onclick="copyCode(this)" style="background-color: #
|
| 87 |
</div>
|
| 88 |
-
<pre
|
| 89 |
pip install accelerate</pre>
|
| 90 |
</div>
|
| 91 |
|
| 92 |
<div style="background-color: #000000; border: 1px solid #333333; border-radius: 4px; margin-bottom: 16px; overflow: hidden;">
|
| 93 |
-
<div style="background-color: #
|
| 94 |
-
<span>python</span>
|
| 95 |
-
<button onclick="copyCode(this)" style="background-color: #
|
| 96 |
</div>
|
| 97 |
-
<pre
|
| 98 |
import torch
|
| 99 |
|
| 100 |
model = Qwen3_5ForConditionalGeneration.from_pretrained(
|
|
@@ -149,73 +148,69 @@ print(output_text[0])</pre>
|
|
| 149 |
</div>
|
| 150 |
|
| 151 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Training Details</h2>
|
| 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 |
-
</tr>
|
| 192 |
-
</tbody>
|
| 193 |
-
</table>
|
| 194 |
-
</div>
|
| 195 |
|
| 196 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Intended Use</h2>
|
| 197 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 198 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 199 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 200 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 201 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 202 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 203 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 204 |
|
| 205 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Limitations</h2>
|
| 206 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 207 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 208 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 209 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 210 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 211 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 212 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 213 |
|
| 214 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Acknowledgements</h2>
|
| 215 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 216 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 217 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 218 |
-
<div style="display: flex; align-items: flex-start; margin-bottom:
|
| 219 |
|
| 220 |
</div>
|
| 221 |
|
|
@@ -229,8 +224,12 @@ print(output_text[0])</pre>
|
|
| 229 |
try {
|
| 230 |
document.execCommand('copy');
|
| 231 |
btn.innerText = 'Copied!';
|
|
|
|
|
|
|
| 232 |
setTimeout(() => {
|
| 233 |
btn.innerText = 'Copy';
|
|
|
|
|
|
|
| 234 |
}, 2000);
|
| 235 |
} catch (err) {
|
| 236 |
console.error('Failed to copy text: ', err);
|
|
|
|
| 7 |
|
| 8 |
.vg-code, .vg-code * {
|
| 9 |
font-family: 'Ubuntu Mono', 'Courier New', monospace !important;
|
|
|
|
| 10 |
}
|
| 11 |
</style>
|
| 12 |
|
|
|
|
| 24 |
|
| 25 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Key Highlights</h2>
|
| 26 |
|
| 27 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Qwen 3.5 Foundation</b>: Built on <b>Qwen/Qwen3.5-4B</b>.</span></div>
|
| 28 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Multimodal Content Filter</b>: Designed for visual content-safety classification.</span></div>
|
| 29 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>ImageShield-Guardrail-Pro Training</b>: Fine-tuned on the <b>prithivMLmods/ImageShield-Guardrail-Pro</b> content-safety guardrail dataset.</span></div>
|
| 30 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Safe / Unsafe Classification</b>: Classifies visual content as either <b>Safe</b> or <b>Unsafe</b>.</span></div>
|
| 31 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>NSFW Sensual Content Detection</b>: Focused on identifying potentially <b>Not Safe for Work (NSFW) sensual content</b> and related sensitive visual material.</span></div>
|
| 32 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Explicit & Sensitive Content Detection</b>: Excels at identifying explicit, revealing, provocative, and otherwise potentially sensitive visual content.</span></div>
|
| 33 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Dress-Code Analysis</b>: Performs detailed analysis of clothing, dress codes, exposure, and revealing garments.</span></div>
|
| 34 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Pose Analysis</b>: Analyzes body poses, positioning, framing, and potentially suggestive presentation.</span></div>
|
| 35 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Context & Setting Analysis</b>: Considers different visual settings and contextual cues when determining content safety.</span></div>
|
| 36 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Zero-Tolerance Safety Classification</b>: Uses a conservative classification policy that favors safety when the content is ambiguous.</span></div>
|
| 37 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Structured Output</b>: Produces a machine-readable JSON response containing a caption, classification, and reasoning.</span></div>
|
| 38 |
|
| 39 |
<details style="border: 1px solid #333333; border-radius: 4px; margin-bottom: 16px; overflow: hidden; background-color: #000000;">
|
| 40 |
+
<summary style="padding: 14px 16px; cursor: pointer; color: #FFFFFF; font-weight: 500; background-color: #111111; list-style: none; display: flex; justify-content: space-between; align-items: center; font-family: 'Ubuntu', sans-serif;">Standard Instruction</summary>
|
| 41 |
+
<div style="background-color: #000000; border: none; margin: 0; overflow: hidden;">
|
| 42 |
+
<div style="background-color: #111111; padding: 8px 16px; border-bottom: 1px solid #333333; display: flex; justify-content: space-between; align-items: center;">
|
| 43 |
+
<span style="font-size: 13px; color: #888888; font-family: 'Ubuntu', sans-serif;">text</span>
|
| 44 |
+
<button onclick="copyCode(this)" style="background-color: #222222; color: #FFFFFF; border: 1px solid #444444; padding: 4px 12px; cursor: pointer; border-radius: 4px; font-size: 12px; font-family: 'Ubuntu', sans-serif;">Copy</button>
|
| 45 |
</div>
|
| 46 |
+
<pre style="margin: 0; padding: 16px; background-color: #000000 !important; color: #FFFFFF !important; white-space: pre-wrap; overflow-x: auto; font-family: 'Ubuntu Mono', 'Courier New', monospace !important; font-size: 14px; line-height: 1.5;">You are a strict, zero-tolerance parental control content-safety guardrail classifier for visual media.
|
| 47 |
|
| 48 |
Your role is minor protection: any media containing suggestive, revealing, sexually provocative, or explicit visual elements must be flagged as NSFW.
|
| 49 |
|
|
|
|
| 80 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Quick Start with Transformers</h2>
|
| 81 |
|
| 82 |
<div style="background-color: #000000; border: 1px solid #333333; border-radius: 4px; margin-bottom: 16px; overflow: hidden;">
|
| 83 |
+
<div style="background-color: #111111; padding: 8px 16px; border-bottom: 1px solid #333333; display: flex; justify-content: space-between; align-items: center;">
|
| 84 |
+
<span style="font-size: 13px; color: #888888; font-family: 'Ubuntu', sans-serif;">bash</span>
|
| 85 |
+
<button onclick="copyCode(this)" style="background-color: #222222; color: #FFFFFF; border: 1px solid #444444; padding: 4px 12px; cursor: pointer; border-radius: 4px; font-size: 12px; font-family: 'Ubuntu', sans-serif;">Copy</button>
|
| 86 |
</div>
|
| 87 |
+
<pre style="margin: 0; padding: 16px; background-color: #000000 !important; color: #FFFFFF !important; white-space: pre-wrap; overflow-x: auto; font-family: 'Ubuntu Mono', 'Courier New', monospace !important; font-size: 14px; line-height: 1.5;">pip install transformers
|
| 88 |
pip install accelerate</pre>
|
| 89 |
</div>
|
| 90 |
|
| 91 |
<div style="background-color: #000000; border: 1px solid #333333; border-radius: 4px; margin-bottom: 16px; overflow: hidden;">
|
| 92 |
+
<div style="background-color: #111111; padding: 8px 16px; border-bottom: 1px solid #333333; display: flex; justify-content: space-between; align-items: center;">
|
| 93 |
+
<span style="font-size: 13px; color: #888888; font-family: 'Ubuntu', sans-serif;">python</span>
|
| 94 |
+
<button onclick="copyCode(this)" style="background-color: #222222; color: #FFFFFF; border: 1px solid #444444; padding: 4px 12px; cursor: pointer; border-radius: 4px; font-size: 12px; font-family: 'Ubuntu', sans-serif;">Copy</button>
|
| 95 |
</div>
|
| 96 |
+
<pre style="margin: 0; padding: 16px; background-color: #000000 !important; color: #FFFFFF !important; white-space: pre-wrap; overflow-x: auto; font-family: 'Ubuntu Mono', 'Courier New', monospace !important; font-size: 14px; line-height: 1.5;">from transformers import Qwen3_5ForConditionalGeneration, AutoProcessor
|
| 97 |
import torch
|
| 98 |
|
| 99 |
model = Qwen3_5ForConditionalGeneration.from_pretrained(
|
|
|
|
| 148 |
</div>
|
| 149 |
|
| 150 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Training Details</h2>
|
| 151 |
+
|
| 152 |
+
<table style="width: 100%; border-collapse: collapse; margin: 24px 0; color: #FFFFFF; background-color: #000000; border: 1px solid #333333; border-radius: 4px; overflow: hidden; font-family: 'Ubuntu', sans-serif;">
|
| 153 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 154 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; width: 40%; border-right: 1px solid #333333; background-color: #000000;">Base Model</td>
|
| 155 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #000000;"><b>Qwen/Qwen3.5-4B</b></td>
|
| 156 |
+
</tr>
|
| 157 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 158 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #0a0a0a;">Model</td>
|
| 159 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #0a0a0a;"><b>VisionGuardrail-4B</b></td>
|
| 160 |
+
</tr>
|
| 161 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 162 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #000000;">Model Type</td>
|
| 163 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #000000;"><b>Multimodal Content-Safety Classifier</b></td>
|
| 164 |
+
</tr>
|
| 165 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 166 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #0a0a0a;">Training Dataset</td>
|
| 167 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #0a0a0a;"><b>prithivMLmods/ImageShield-Guardrail-Pro</b></td>
|
| 168 |
+
</tr>
|
| 169 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 170 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #000000;">Training Objective</td>
|
| 171 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #000000;"><b>Content-safety guardrail classification</b></td>
|
| 172 |
+
</tr>
|
| 173 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 174 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #0a0a0a;">Primary Classification</td>
|
| 175 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #0a0a0a;"><b>Safe / Unsafe</b></td>
|
| 176 |
+
</tr>
|
| 177 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 178 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #000000;">Safety Focus</td>
|
| 179 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #000000;"><b>Not Safe for Work (NSFW) Sensual Content</b></td>
|
| 180 |
+
</tr>
|
| 181 |
+
<tr style="border-bottom: 1px solid #333333;">
|
| 182 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #0a0a0a;">Visual Analysis</td>
|
| 183 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #0a0a0a;"><b>Dress Codes, Poses, Exposure & Settings</b></td>
|
| 184 |
+
</tr>
|
| 185 |
+
<tr>
|
| 186 |
+
<td style="padding: 14px 16px; color: #AAAAAA; font-weight: 500; border-right: 1px solid #333333; background-color: #000000;">Training Framework</td>
|
| 187 |
+
<td style="padding: 14px 16px; color: #FFFFFF; background-color: #000000;"><b>TRL</b></td>
|
| 188 |
+
</tr>
|
| 189 |
+
</table>
|
|
|
|
|
|
|
|
|
|
|
|
|
| 190 |
|
| 191 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Intended Use</h2>
|
| 192 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Content Safety Classification</b>: Classifying visual media as safe or unsafe.</span></div>
|
| 193 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>NSFW Sensual Content Detection</b>: Supporting research into automated detection of potentially sensitive and sexually suggestive visual content.</span></div>
|
| 194 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Parental Controls</b>: Building conservative visual content-safety filtering systems.</span></div>
|
| 195 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Content Moderation</b>: Supporting automated safety classification pipelines.</span></div>
|
| 196 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Multimodal Safety Research</b>: Evaluating content-safety behavior in multimodal language models.</span></div>
|
| 197 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Guardrail Development</b>: Researching structured safety classification and filtering workflows.</span></div>
|
| 198 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Visual Safety Analysis</b>: Analyzing clothing, dress codes, poses, framing, and visual settings for content-safety signals.</span></div>
|
| 199 |
|
| 200 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Limitations</h2>
|
| 201 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Experimental Model</b>: The model may produce incorrect or inconsistent classifications.</span></div>
|
| 202 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>False Positives</b>: Benign content may occasionally be classified as unsafe due to the conservative classification threshold.</span></div>
|
| 203 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>False Negatives</b>: Unsafe content may occasionally be missed.</span></div>
|
| 204 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Context Sensitivity</b>: Classification performance depends on image quality, visual context, and the provided instruction.</span></div>
|
| 205 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Conservative Policy</b>: The model intentionally uses a strict classification threshold and may flag content that would not be considered NSFW under less restrictive moderation policies.</span></div>
|
| 206 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Visual Ambiguity</b>: Clothing, poses, artistic styles, and contextual cues may sometimes be difficult to interpret accurately.</span></div>
|
| 207 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>Automated Classification</b>: The model should not be treated as a definitive legal or safety determination.</span></div>
|
| 208 |
|
| 209 |
<h2 style="color: #FFFFFF; font-weight: 500; border-bottom: 1px solid #333333; padding-bottom: 8px; margin-top: 32px;">Acknowledgements</h2>
|
| 210 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><a href="https://huggingface.co/Qwen/Qwen3.5-4B" style="color: #FF4500; text-decoration: none;">Qwen/Qwen3.5-4B</a>: Base multimodal model used for this project.</span></div>
|
| 211 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><a href="https://huggingface.co/datasets/prithivMLmods/ImageShield-Guardrail-Pro" style="color: #FF4500; text-decoration: none;">ImageShield-Guardrail-Pro</a>: Content-safety guardrail dataset used for training VisionGuardrail-4B.</span></div>
|
| 212 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><b>TRL</b> – <a href="https://huggingface.co/docs/trl/en/index" style="color: #FF4500; text-decoration: none;">Transformers Reinforcement Learning</a>: TRL is a full-stack library providing tools to train transformer language models with methods including Supervised Fine-Tuning (SFT), Group Relative Policy Optimization (GRPO), Direct Preference Optimization (DPO), Reward Modeling, and more.</span></div>
|
| 213 |
+
<div style="display: flex; align-items: flex-start; margin-bottom: 10px; color: #FFFFFF;"><span style="margin-right: 12px; color: #FFFFFF; font-weight: bold; line-height: 1.5;">+</span><span style="line-height: 1.5;"><a href="https://huggingface.co/docs/transformers/en/index" style="color: #FF4500; text-decoration: none;">Transformers</a>: Transformers provides state-of-the-art machine learning models for text, computer vision, audio, video, and multimodal tasks, supporting both inference and training.</span></div>
|
| 214 |
|
| 215 |
</div>
|
| 216 |
|
|
|
|
| 224 |
try {
|
| 225 |
document.execCommand('copy');
|
| 226 |
btn.innerText = 'Copied!';
|
| 227 |
+
btn.style.backgroundColor = '#FF4500';
|
| 228 |
+
btn.style.borderColor = '#FF4500';
|
| 229 |
setTimeout(() => {
|
| 230 |
btn.innerText = 'Copy';
|
| 231 |
+
btn.style.backgroundColor = '#222222';
|
| 232 |
+
btn.style.borderColor = '#444444';
|
| 233 |
}, 2000);
|
| 234 |
} catch (err) {
|
| 235 |
console.error('Failed to copy text: ', err);
|