Add slide and rotate interactive captcha solvers

New solver subsystem with independent models:
- GapDetectorCNN (1x128x256 grayscale → sigmoid) for slide gap detection
- RotationRegressor (3x128x128 RGB → sin/cos via tanh) for rotation angle prediction
- SlideSolver with 3-tier strategy: template match → edge detect → CNN fallback
- RotateSolver with ONNX sin/cos → atan2 inference
- Generators, training scripts, CLI commands, and slide track utility

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
Hua
2026-03-11 18:07:06 +08:00
parent 90d6423551
commit 9b5f29083e
20 changed files with 1440 additions and 10 deletions

112
generators/slide_gen.py Normal file
View File

@@ -0,0 +1,112 @@
"""
滑块验证码数据生成器
生成滑块验证码训练数据:随机纹理/色块背景 + 方形缺口 + 阴影效果。
标签 = 缺口中心 x 坐标 (整数)
文件名格式: {gap_x}_{index:06d}.png
"""
import random
from PIL import Image, ImageDraw, ImageFilter
from config import SOLVER_CONFIG
from generators.base import BaseCaptchaGenerator
class SlideDataGenerator(BaseCaptchaGenerator):
"""滑块验证码数据生成器。"""
def __init__(self, seed: int | None = None):
from config import RANDOM_SEED
super().__init__(seed=seed if seed is not None else RANDOM_SEED)
self.cfg = SOLVER_CONFIG["slide"]
self.height, self.width = self.cfg["cnn_input_size"] # (H, W)
self.gap_size = 40 # 缺口大小
def generate(self, text: str | None = None) -> tuple[Image.Image, str]:
rng = self.rng
gs = self.gap_size
# 缺口 x 范围: 留出边距
margin = gs + 10
gap_x = rng.randint(margin, self.width - margin)
gap_y = rng.randint(10, self.height - gs - 10)
if text is None:
text = str(gap_x)
# 1. 生成纹理背景
img = self._textured_background(rng)
# 2. 绘制缺口 (半透明灰色区域 + 阴影)
overlay = Image.new("RGBA", img.size, (0, 0, 0, 0))
overlay_draw = ImageDraw.Draw(overlay)
# 阴影 (稍大一圈)
overlay_draw.rectangle(
[gap_x + 2, gap_y + 2, gap_x + gs + 2, gap_y + gs + 2],
fill=(0, 0, 0, 60),
)
# 缺口本体
overlay_draw.rectangle(
[gap_x, gap_y, gap_x + gs, gap_y + gs],
fill=(80, 80, 80, 160),
outline=(60, 60, 60, 200),
width=2,
)
img = img.convert("RGBA")
img = Image.alpha_composite(img, overlay)
img = img.convert("RGB")
# 3. 轻微模糊
img = img.filter(ImageFilter.GaussianBlur(radius=0.3))
return img, text
def _textured_background(self, rng: random.Random) -> Image.Image:
"""生成带纹理的彩色背景。"""
img = Image.new("RGB", (self.width, self.height))
draw = ImageDraw.Draw(img)
# 渐变底色
base_r = rng.randint(80, 200)
base_g = rng.randint(80, 200)
base_b = rng.randint(80, 200)
for y in range(self.height):
ratio = y / max(self.height - 1, 1)
r = int(base_r + 40 * ratio)
g = int(base_g - 20 * ratio)
b = int(base_b + 20 * ratio)
r, g, b = max(0, min(255, r)), max(0, min(255, g)), max(0, min(255, b))
draw.line([(0, y), (self.width, y)], fill=(r, g, b))
# 纹理噪声
for _ in range(self.width * self.height // 6):
x = rng.randint(0, self.width - 1)
y = rng.randint(0, self.height - 1)
pixel = img.getpixel((x, y))
noise = tuple(
max(0, min(255, c + rng.randint(-30, 30)))
for c in pixel
)
draw.point((x, y), fill=noise)
# 随机色块 (模拟图案)
for _ in range(rng.randint(4, 8)):
x1, y1 = rng.randint(0, self.width - 30), rng.randint(0, self.height - 20)
x2, y2 = x1 + rng.randint(15, 50), y1 + rng.randint(10, 30)
color = tuple(rng.randint(50, 230) for _ in range(3))
draw.rectangle([x1, y1, x2, y2], fill=color)
# 随机圆形
for _ in range(rng.randint(2, 5)):
cx = rng.randint(10, self.width - 10)
cy = rng.randint(10, self.height - 10)
cr = rng.randint(5, 20)
color = tuple(rng.randint(50, 230) for _ in range(3))
draw.ellipse([cx - cr, cy - cr, cx + cr, cy + cr], fill=color)
return img