filter draw plot script

filter.py

from matplotlib import ticker
import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

QPS = [13.42, 12.32, 11.23, 8.75, 27.65, 25.36, 20.41, 16.66]

original_data = {
    "acc": [
        0.9265,
        0.9323,
        0.9533,
        0.9653,
        0.9265,
        0.9323,
        0.9534,
        0.9652
    ],
    "QPS": QPS,
    "": ["baseline"] * 4
    + ["prefilter enabled"] * 4
}


df = pd.DataFrame(original_data)

categories = df[""].unique()

df_curve = pd.DataFrame()

for category in categories:
    category_data = df[df[""] == category]
    acc_min = min(category_data["acc"])
    acc_max = max(category_data["acc"])
    acc_values = np.linspace(acc_min, acc_max, 100)
    rps_curve = np.interp(acc_values, category_data["acc"], category_data["QPS"])

    df_temp = pd.DataFrame({"acc": acc_values, "QPS": rps_curve, "": category})

    df_curve = pd.concat([df_curve, df_temp], ignore_index=True)

# 设置绘图风格和字体
plt.style.use("seaborn-v0_8-whitegrid")
plt.rcParams["font.family"] = "serif"
plt.rcParams["font.size"] = 8

# 创建图形
fig, ax = plt.subplots(figsize=(5, 3))

# 设置背景颜色
ax.set_facecolor("#f0f7f4")
fig.set_facecolor("#f0f7f4")

# 绘制平滑曲线
sns.lineplot(
    x="acc",
    y="QPS",
    hue="",
    data=df_curve,
    dashes=False,
    linewidth=1.5,
    ax=ax,
    palette=["#BF7E04", "#2ca02c"],  # 指定颜色
    legend=False,
)

# 绘制原始数据点
sns.scatterplot(
    x="acc",
    y="QPS",
    hue="",
    data=df,
    style="",
    markers=["o", "o"],
    s=50,
    linewidth=1.5,
    ax=ax,
    palette=["#BF7E04", "#2ca02c"],  # 指定颜色
    legend=True,
)

# 调整图例

# 调整图例

# 添加标题和标签
ax.set_title(
    "LAION-5m Query Per Second (Filter rate 0.01)",
    fontsize=12,
    fontweight="bold",
    pad=20,
)
ax.set_xlabel("Precision@Top 100", fontsize=10)
ax.set_ylabel("QPS", fontsize=10)

# 调整网格线
ax.grid(True, linestyle="--", alpha=0.6)

# 保存图形
plt.savefig("filter.png", dpi=300, bbox_inches="tight")