Spaces:
Running
on
L4
Running
on
L4
Commit
·
d6ff06e
1
Parent(s):
6afc369
Started Space
Browse files- .gitattributes +2 -0
- app.py +177 -0
- examples.json +51 -0
- images/image_0.png +3 -0
- images/image_1.png +3 -0
- images/image_2.png +3 -0
- images/image_3.png +3 -0
- images/image_4.png +3 -0
- images/image_5.png +3 -0
- images/image_6.png +3 -0
- images/original_image_0.png +3 -0
- images/original_image_1.png +3 -0
- images/original_image_2.png +3 -0
- images/original_image_3.png +3 -0
- images/original_image_4.png +3 -0
- images/original_image_5.png +3 -0
- images/original_image_6.png +3 -0
- requirements.txt +8 -0
- utils.py +148 -0
.gitattributes
CHANGED
|
@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
|
|
|
| 33 |
*.zip filter=lfs diff=lfs merge=lfs -text
|
| 34 |
*.zst filter=lfs diff=lfs merge=lfs -text
|
| 35 |
*tfevents* filter=lfs diff=lfs merge=lfs -text
|
| 36 |
+
*.png filter=lfs diff=lfs merge=lfs -text
|
| 37 |
+
*.jpg filter=lfs diff=lfs merge=lfs -text
|
app.py
ADDED
|
@@ -0,0 +1,177 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import torch
|
| 3 |
+
import spaces
|
| 4 |
+
import json
|
| 5 |
+
import base64
|
| 6 |
+
from io import BytesIO
|
| 7 |
+
from transformers import SamHQModel, SamHQProcessor, SamModel, SamProcessor
|
| 8 |
+
import os
|
| 9 |
+
import pandas as pd
|
| 10 |
+
from utils import *
|
| 11 |
+
from PIL import Image
|
| 12 |
+
|
| 13 |
+
# Carga de modelos
|
| 14 |
+
sam_hq_model = SamHQModel.from_pretrained("syscv-community/sam-hq-vit-huge")
|
| 15 |
+
sam_hq_processor = SamHQProcessor.from_pretrained("syscv-community/sam-hq-vit-huge")
|
| 16 |
+
|
| 17 |
+
sam_model = SamModel.from_pretrained("facebook/sam-vit-huge")
|
| 18 |
+
sam_processor = SamProcessor.from_pretrained("facebook/sam-vit-huge")
|
| 19 |
+
|
| 20 |
+
@spaces.GPU
|
| 21 |
+
def predict_masks_and_scores(model, processor, raw_image, input_points=None, input_boxes=None):
|
| 22 |
+
if input_boxes is not None:
|
| 23 |
+
input_boxes = [input_boxes]
|
| 24 |
+
inputs = processor(raw_image, input_boxes=input_boxes, input_points=input_points, return_tensors="pt")
|
| 25 |
+
with torch.no_grad():
|
| 26 |
+
outputs = model(**inputs)
|
| 27 |
+
|
| 28 |
+
masks = processor.image_processor.post_process_masks(
|
| 29 |
+
outputs.pred_masks.cpu(), inputs["original_sizes"].cpu(), inputs["reshaped_input_sizes"].cpu()
|
| 30 |
+
)
|
| 31 |
+
scores = outputs.iou_scores
|
| 32 |
+
return masks, scores
|
| 33 |
+
|
| 34 |
+
def encode_pil_to_base64(pil_image):
|
| 35 |
+
buffer = BytesIO()
|
| 36 |
+
pil_image.save(buffer, format="PNG")
|
| 37 |
+
return base64.b64encode(buffer.getvalue()).decode("utf-8")
|
| 38 |
+
|
| 39 |
+
def compare_images_points_and_masks(user_image, input_boxes, input_points):
|
| 40 |
+
for example_path, example_data in example_data_map.items():
|
| 41 |
+
if example_data["size"] == list(user_image.size):
|
| 42 |
+
user_image = Image.open(example_data['original_image_path'])
|
| 43 |
+
input_boxes = input_boxes.values.tolist()
|
| 44 |
+
input_points = input_points.values.tolist()
|
| 45 |
+
|
| 46 |
+
input_boxes = [[[int(coord) for coord in box] for box in input_boxes if any(box)]]
|
| 47 |
+
input_points = [[[int(coord) for coord in point] for point in input_points if any(point)]]
|
| 48 |
+
|
| 49 |
+
input_boxes = input_boxes if input_boxes[0] else None
|
| 50 |
+
input_points = input_points if input_points[0] else None
|
| 51 |
+
|
| 52 |
+
sam_masks, sam_scores = predict_masks_and_scores(sam_model, sam_processor, user_image, input_boxes=input_boxes, input_points=input_points)
|
| 53 |
+
sam_hq_masks, sam_hq_scores = predict_masks_and_scores(sam_hq_model, sam_hq_processor, user_image, input_boxes=input_boxes, input_points=input_points)
|
| 54 |
+
|
| 55 |
+
if input_boxes and input_points:
|
| 56 |
+
img1_b64 = show_all_annotations_on_image_base64(user_image, sam_masks[0][0], sam_scores[:, 0, :], input_boxes[0], input_points[0], model_name='SAM')
|
| 57 |
+
img2_b64 = show_all_annotations_on_image_base64(user_image, sam_hq_masks[0][0], sam_hq_scores[:, 0, :], input_boxes[0], input_points[0], model_name='SAM_HQ')
|
| 58 |
+
elif input_boxes:
|
| 59 |
+
img1_b64 = show_all_annotations_on_image_base64(user_image, sam_masks[0][0], sam_scores[:, 0, :], input_boxes[0], None, model_name='SAM')
|
| 60 |
+
img2_b64 = show_all_annotations_on_image_base64(user_image, sam_hq_masks[0][0], sam_hq_scores[:, 0, :], input_boxes[0], None, model_name='SAM_HQ')
|
| 61 |
+
elif input_points:
|
| 62 |
+
img1_b64 = show_all_annotations_on_image_base64(user_image, sam_masks[0][0], sam_scores[:, 0, :], None, input_points[0], model_name='SAM')
|
| 63 |
+
img2_b64 = show_all_annotations_on_image_base64(user_image, sam_hq_masks[0][0], sam_hq_scores[:, 0, :], None, input_points[0], model_name='SAM_HQ')
|
| 64 |
+
|
| 65 |
+
print('user_image', user_image)
|
| 66 |
+
print("img1_b64", img1_b64)
|
| 67 |
+
print("img2_b64", img2_b64)
|
| 68 |
+
|
| 69 |
+
html_code = f"""
|
| 70 |
+
<div style="position: relative; width: 100%; max-width: 600px; margin: 0 auto;" id="imageCompareContainer">
|
| 71 |
+
<div style="position: relative; width: 100%;">
|
| 72 |
+
<img src="data:image/png;base64,{img1_b64}" style="width:100%; display:block;">
|
| 73 |
+
<div id="topWrapper" style="position:absolute; top:0; left:0; width:100%; overflow:hidden;">
|
| 74 |
+
<img id="topImage" src="data:image/png;base64,{img2_b64}" style="width:100%;">
|
| 75 |
+
</div>
|
| 76 |
+
<div id="sliderLine" style="position:absolute; top:0; left:0; width:2px; height:100%; background-color:red; pointer-events:none;"></div>
|
| 77 |
+
</div>
|
| 78 |
+
<input type="range" min="0" max="100" value="0"
|
| 79 |
+
style="width:100%; margin-top: 10px;"
|
| 80 |
+
oninput="
|
| 81 |
+
const val = this.value;
|
| 82 |
+
const container = document.getElementById('imageCompareContainer');
|
| 83 |
+
const width = container.offsetWidth;
|
| 84 |
+
const clipValue = 100 - val;
|
| 85 |
+
document.getElementById('topImage').style.clipPath = 'inset(0 ' + clipValue + '% 0 0)';
|
| 86 |
+
document.getElementById('sliderLine').style.left = (width * val / 100) + 'px';
|
| 87 |
+
">
|
| 88 |
+
</div>
|
| 89 |
+
"""
|
| 90 |
+
return html_code
|
| 91 |
+
|
| 92 |
+
def load_examples(json_file="examples.json"):
|
| 93 |
+
with open(json_file, "r") as f:
|
| 94 |
+
examples = json.load(f)
|
| 95 |
+
return examples
|
| 96 |
+
|
| 97 |
+
examples = load_examples()
|
| 98 |
+
example_paths = [example["image_path"] for example in examples]
|
| 99 |
+
example_data_map = {
|
| 100 |
+
example["image_path"]: {
|
| 101 |
+
"original_image_path": example["original_image_path"],
|
| 102 |
+
"points": example["points"],
|
| 103 |
+
"boxes": example["boxes"],
|
| 104 |
+
"size": example["size"]
|
| 105 |
+
}
|
| 106 |
+
for example in examples
|
| 107 |
+
}
|
| 108 |
+
|
| 109 |
+
theme = gr.themes.Soft(primary_hue="indigo", secondary_hue="emerald")
|
| 110 |
+
with gr.Blocks(theme=theme, title="🔍 Compare SAM vs SAM-HQ") as demo:
|
| 111 |
+
image_path_box = gr.Textbox(visible=False)
|
| 112 |
+
gr.Markdown("## 🔍 Compare SAM vs SAM-HQ")
|
| 113 |
+
gr.Markdown("Compare the performance of SAM and SAM-HQ on various images. Click on an example to load it")
|
| 114 |
+
gr.Markdown("[SAM-HQ](https://huggingface.co/syscv-community/sam-hq-vit-huge) - [SAM](https://huggingface.co/facebook/sam-vit-huge)")
|
| 115 |
+
|
| 116 |
+
with gr.Row():
|
| 117 |
+
image_input = gr.Image(
|
| 118 |
+
type="pil",
|
| 119 |
+
label="Example image (click below to load)",
|
| 120 |
+
interactive=False,
|
| 121 |
+
height=500,
|
| 122 |
+
show_label=True
|
| 123 |
+
)
|
| 124 |
+
|
| 125 |
+
gr.Examples(
|
| 126 |
+
examples=example_paths,
|
| 127 |
+
inputs=[image_input],
|
| 128 |
+
label="Click an example to try 👇",
|
| 129 |
+
)
|
| 130 |
+
|
| 131 |
+
result_html = gr.HTML(elem_id="result-html")
|
| 132 |
+
|
| 133 |
+
with gr.Row():
|
| 134 |
+
points_input = gr.Dataframe(
|
| 135 |
+
headers=["x", "y"],
|
| 136 |
+
label="Points",
|
| 137 |
+
datatype=["number", "number"],
|
| 138 |
+
col_count=(2, "fixed")
|
| 139 |
+
)
|
| 140 |
+
boxes_input = gr.Dataframe(
|
| 141 |
+
headers=["x0", "y0", "x1", "y1"],
|
| 142 |
+
label="Boxes",
|
| 143 |
+
datatype=["number", "number", "number", "number"],
|
| 144 |
+
col_count=(4, "fixed")
|
| 145 |
+
)
|
| 146 |
+
|
| 147 |
+
def on_image_change(image):
|
| 148 |
+
for example_path, example_data in example_data_map.items():
|
| 149 |
+
print(image.size)
|
| 150 |
+
if example_data["size"] == list(image.size):
|
| 151 |
+
return example_data["points"], example_data["boxes"]
|
| 152 |
+
return [], []
|
| 153 |
+
|
| 154 |
+
image_input.change(
|
| 155 |
+
fn=on_image_change,
|
| 156 |
+
inputs=[image_input],
|
| 157 |
+
outputs=[points_input, boxes_input]
|
| 158 |
+
)
|
| 159 |
+
|
| 160 |
+
compare_button = gr.Button("Compare points and masks")
|
| 161 |
+
compare_button.click(fn=compare_images_points_and_masks, inputs=[image_input, boxes_input, points_input], outputs=result_html)
|
| 162 |
+
|
| 163 |
+
gr.HTML("""
|
| 164 |
+
<style>
|
| 165 |
+
#result-html {
|
| 166 |
+
min-height: 500px;
|
| 167 |
+
border: 1px solid #ccc;
|
| 168 |
+
padding: 10px;
|
| 169 |
+
box-sizing: border-box;
|
| 170 |
+
background-color: #fff;
|
| 171 |
+
border-radius: 8px;
|
| 172 |
+
box-shadow: 0 2px 6px rgba(0, 0, 0, 0.1);
|
| 173 |
+
}
|
| 174 |
+
</style>
|
| 175 |
+
""")
|
| 176 |
+
|
| 177 |
+
demo.launch()
|
examples.json
ADDED
|
@@ -0,0 +1,51 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[
|
| 2 |
+
{
|
| 3 |
+
"image_path": "./images/image_0.png",
|
| 4 |
+
"original_image_path": "./images/original_image_0.png",
|
| 5 |
+
"points": null,
|
| 6 |
+
"boxes": [[4,13,1007,1023]],
|
| 7 |
+
"size": [1024, 1024]
|
| 8 |
+
},
|
| 9 |
+
{
|
| 10 |
+
"image_path": "./images/image_1.png",
|
| 11 |
+
"original_image_path": "./images/original_image_1.png",
|
| 12 |
+
"points": null,
|
| 13 |
+
"boxes": [[230, 99, 694, 670]],
|
| 14 |
+
"size": [768, 768]
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"image_path": "./images/image_2.png",
|
| 18 |
+
"original_image_path": "./images/original_image_2.png",
|
| 19 |
+
"points": [[495,518],[217,140]],
|
| 20 |
+
"boxes": null,
|
| 21 |
+
"size": [894, 1000]
|
| 22 |
+
},
|
| 23 |
+
{
|
| 24 |
+
"image_path": "./images/image_3.png",
|
| 25 |
+
"original_image_path": "./images/original_image_3.png",
|
| 26 |
+
"points": [[111, 241],[249, 317],[375, 190]],
|
| 27 |
+
"boxes": null,
|
| 28 |
+
"size": [512, 512]
|
| 29 |
+
},
|
| 30 |
+
{
|
| 31 |
+
"image_path": "./images/image_4.png",
|
| 32 |
+
"original_image_path": "./images/original_image_4.png",
|
| 33 |
+
"points": null,
|
| 34 |
+
"boxes": [[128, 152, 1880, 1838]],
|
| 35 |
+
"size": [2048, 2048]
|
| 36 |
+
},
|
| 37 |
+
{
|
| 38 |
+
"image_path": "./images/image_5.png",
|
| 39 |
+
"original_image_path": "./images/original_image_5.png",
|
| 40 |
+
"points": [[373,363], [452, 575]],
|
| 41 |
+
"boxes": null,
|
| 42 |
+
"size": [1024, 683]
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"image_path": "./images/image_6.png",
|
| 46 |
+
"original_image_path": "./images/original_image_6.png",
|
| 47 |
+
"points": null,
|
| 48 |
+
"boxes": [[181, 196, 757, 495]],
|
| 49 |
+
"size": [800, 533]
|
| 50 |
+
}
|
| 51 |
+
]
|
images/image_0.png
ADDED
|
Git LFS Details
|
images/image_1.png
ADDED
|
Git LFS Details
|
images/image_2.png
ADDED
|
Git LFS Details
|
images/image_3.png
ADDED
|
Git LFS Details
|
images/image_4.png
ADDED
|
Git LFS Details
|
images/image_5.png
ADDED
|
Git LFS Details
|
images/image_6.png
ADDED
|
Git LFS Details
|
images/original_image_0.png
ADDED
|
Git LFS Details
|
images/original_image_1.png
ADDED
|
Git LFS Details
|
images/original_image_2.png
ADDED
|
Git LFS Details
|
images/original_image_3.png
ADDED
|
Git LFS Details
|
images/original_image_4.png
ADDED
|
Git LFS Details
|
images/original_image_5.png
ADDED
|
Git LFS Details
|
images/original_image_6.png
ADDED
|
Git LFS Details
|
requirements.txt
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
gradio
|
| 2 |
+
huggingface_hub
|
| 3 |
+
requests
|
| 4 |
+
pillow
|
| 5 |
+
torch
|
| 6 |
+
git+https://github.com/huggingface/transformers.git
|
| 7 |
+
matplotlib
|
| 8 |
+
numpy
|
utils.py
ADDED
|
@@ -0,0 +1,148 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
from io import BytesIO
|
| 2 |
+
import base64
|
| 3 |
+
import numpy as np
|
| 4 |
+
import matplotlib.pyplot as plt
|
| 5 |
+
import torch
|
| 6 |
+
|
| 7 |
+
|
| 8 |
+
def fig_to_base64(fig):
|
| 9 |
+
buf = BytesIO()
|
| 10 |
+
fig.savefig(buf, format='png', bbox_inches='tight')
|
| 11 |
+
plt.close(fig)
|
| 12 |
+
buf.seek(0)
|
| 13 |
+
return base64.b64encode(buf.getvalue()).decode()
|
| 14 |
+
|
| 15 |
+
def show_mask(mask, ax, random_color=False):
|
| 16 |
+
if random_color:
|
| 17 |
+
color = np.concatenate([np.random.random(3), np.array([0.6])], axis=0)
|
| 18 |
+
else:
|
| 19 |
+
color = np.array([30/255, 144/255, 255/255, 0.6])
|
| 20 |
+
h, w = mask.shape[-2:]
|
| 21 |
+
mask_image = mask.reshape(h, w, 1) * color.reshape(1, 1, -1)
|
| 22 |
+
ax.imshow(mask_image)
|
| 23 |
+
|
| 24 |
+
def show_box(box, ax):
|
| 25 |
+
x0, y0 = box[0], box[1]
|
| 26 |
+
w, h = box[2] - box[0], box[3] - box[1]
|
| 27 |
+
ax.add_patch(plt.Rectangle((x0, y0), w, h, edgecolor='green', facecolor=(0,0,0,0), lw=2))
|
| 28 |
+
|
| 29 |
+
def show_points(coords, labels, ax, marker_size=375):
|
| 30 |
+
pos_points = coords[labels==1]
|
| 31 |
+
neg_points = coords[labels==0]
|
| 32 |
+
ax.scatter(pos_points[:, 0], pos_points[:, 1], color='green', marker='*', s=marker_size, edgecolor='white', linewidth=1.25)
|
| 33 |
+
ax.scatter(neg_points[:, 0], neg_points[:, 1], color='red', marker='*', s=marker_size, edgecolor='white', linewidth=1.25)
|
| 34 |
+
|
| 35 |
+
def show_boxes_on_image_base64(raw_image, boxes):
|
| 36 |
+
fig, ax = plt.subplots(figsize=(10,10))
|
| 37 |
+
ax.imshow(raw_image)
|
| 38 |
+
for box in boxes:
|
| 39 |
+
show_box(box, ax)
|
| 40 |
+
ax.axis('off')
|
| 41 |
+
return fig_to_base64(fig)
|
| 42 |
+
|
| 43 |
+
def show_points_on_image_base64(raw_image, input_points, input_labels=None):
|
| 44 |
+
fig, ax = plt.subplots(figsize=(10,10))
|
| 45 |
+
ax.imshow(raw_image)
|
| 46 |
+
input_points = np.array(input_points)
|
| 47 |
+
labels = np.ones_like(input_points[:, 0]) if input_labels is None else np.array(input_labels)
|
| 48 |
+
show_points(input_points, labels, ax)
|
| 49 |
+
ax.axis('off')
|
| 50 |
+
return fig_to_base64(fig)
|
| 51 |
+
|
| 52 |
+
def show_points_and_boxes_on_image_base64(raw_image, boxes, input_points, input_labels=None):
|
| 53 |
+
fig, ax = plt.subplots(figsize=(10,10))
|
| 54 |
+
ax.imshow(raw_image)
|
| 55 |
+
input_points = np.array(input_points)
|
| 56 |
+
labels = np.ones_like(input_points[:, 0]) if input_labels is None else np.array(input_labels)
|
| 57 |
+
show_points(input_points, labels, ax)
|
| 58 |
+
for box in boxes:
|
| 59 |
+
show_box(box, ax)
|
| 60 |
+
ax.axis('off')
|
| 61 |
+
return fig_to_base64(fig)
|
| 62 |
+
|
| 63 |
+
def show_masks_on_image_base64(raw_image, masks, scores):
|
| 64 |
+
if len(masks.shape) == 4:
|
| 65 |
+
masks = masks.squeeze()
|
| 66 |
+
if scores.shape[0] == 1:
|
| 67 |
+
scores = scores.squeeze()
|
| 68 |
+
|
| 69 |
+
nb_predictions = scores.shape[-1]
|
| 70 |
+
print(f"Number of predictions: {nb_predictions}")
|
| 71 |
+
fig, axes = plt.subplots(1, nb_predictions, figsize=(5 * nb_predictions, 5))
|
| 72 |
+
|
| 73 |
+
if nb_predictions == 1:
|
| 74 |
+
axes = [axes]
|
| 75 |
+
|
| 76 |
+
for i, (mask, score) in enumerate(zip(masks, scores)):
|
| 77 |
+
print(i)
|
| 78 |
+
mask = mask.cpu().detach().numpy()
|
| 79 |
+
axes[i].imshow(np.array(raw_image))
|
| 80 |
+
show_mask(mask, axes[i])
|
| 81 |
+
axes[i].title.set_text(f"Mask {i+1}, Score: {score.item():.3f}")
|
| 82 |
+
axes[i].axis("off")
|
| 83 |
+
|
| 84 |
+
return fig_to_base64(fig)
|
| 85 |
+
|
| 86 |
+
def show_first_mask_on_image_base64(raw_image, masks, scores):
|
| 87 |
+
if masks.ndim == 4:
|
| 88 |
+
mask = masks[0, 0]
|
| 89 |
+
elif masks.ndim == 3:
|
| 90 |
+
mask = masks[0]
|
| 91 |
+
else:
|
| 92 |
+
mask = masks
|
| 93 |
+
|
| 94 |
+
if isinstance(mask, torch.Tensor):
|
| 95 |
+
mask = mask.cpu().detach().numpy()
|
| 96 |
+
|
| 97 |
+
score_text = ""
|
| 98 |
+
if scores is not None:
|
| 99 |
+
if isinstance(scores, torch.Tensor):
|
| 100 |
+
scores = scores.flatten()
|
| 101 |
+
score = scores[0].item()
|
| 102 |
+
else:
|
| 103 |
+
score = float(np.array(scores).flatten()[0])
|
| 104 |
+
score_text = f"Score: {score:.3f}"
|
| 105 |
+
|
| 106 |
+
fig, ax = plt.subplots(figsize=(5, 5))
|
| 107 |
+
ax.imshow(np.array(raw_image))
|
| 108 |
+
show_mask(mask, ax)
|
| 109 |
+
ax.set_title(score_text)
|
| 110 |
+
ax.axis("off")
|
| 111 |
+
|
| 112 |
+
return fig_to_base64(fig)
|
| 113 |
+
|
| 114 |
+
def show_all_annotations_on_image_base64(raw_image, masks=None, scores=None, boxes=None, input_points=None, input_labels=None, model_name=None):
|
| 115 |
+
fig, ax = plt.subplots(figsize=(10, 10))
|
| 116 |
+
ax.imshow(np.array(raw_image))
|
| 117 |
+
|
| 118 |
+
if masks is not None:
|
| 119 |
+
if masks.ndim == 4:
|
| 120 |
+
mask = masks[0, 0]
|
| 121 |
+
elif masks.ndim == 3:
|
| 122 |
+
mask = masks[0]
|
| 123 |
+
else:
|
| 124 |
+
mask = masks
|
| 125 |
+
if isinstance(mask, torch.Tensor):
|
| 126 |
+
mask = mask.cpu().detach().numpy()
|
| 127 |
+
show_mask(mask, ax)
|
| 128 |
+
|
| 129 |
+
if scores is not None:
|
| 130 |
+
if isinstance(scores, torch.Tensor):
|
| 131 |
+
scores = scores.flatten()
|
| 132 |
+
score = scores[0].item()
|
| 133 |
+
else:
|
| 134 |
+
score = float(np.array(scores).flatten()[0])
|
| 135 |
+
ax.set_title(f"{model_name} - Score: {score:.3f}")
|
| 136 |
+
|
| 137 |
+
|
| 138 |
+
if input_points is not None:
|
| 139 |
+
input_points = np.array(input_points)
|
| 140 |
+
labels = np.ones_like(input_points[:, 0]) if input_labels is None else np.array(input_labels)
|
| 141 |
+
show_points(input_points, labels, ax)
|
| 142 |
+
|
| 143 |
+
if boxes is not None:
|
| 144 |
+
for box in boxes:
|
| 145 |
+
show_box(box, ax)
|
| 146 |
+
|
| 147 |
+
ax.axis("off")
|
| 148 |
+
return fig_to_base64(fig)
|