Datasets:
Dataset Viewer
The dataset viewer is not available for this subset.
Cannot get the split names for the config 'default' of the dataset.
Exception: SplitsNotFoundError
Message: The split names could not be parsed from the dataset config.
Traceback: Traceback (most recent call last):
File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
for split_generator in builder._split_generators(
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/packaged_modules/webdataset/webdataset.py", line 82, in _split_generators
raise ValueError(
ValueError: The TAR archives of the dataset should be in WebDataset format, but the files in the archive don't share the same prefix or the same types.
The above exception was the direct cause of the following exception:
Traceback (most recent call last):
File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 66, in compute_split_names_from_streaming_response
for split in get_dataset_split_names(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
info = get_dataset_config_info(
^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.12/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
SciEval-NLOS:非视域成像下的语义轮廓复原数据集
本数据集对应项目 NLOS-LDM——基于 STL-10 的「非视域(NLOS)场景采集 → 背景扣除 → 网络复原」实验, 通过深度网络从墙面投影图中初步复原出带有类别(label)信息的目标轮廓图。
- 📂 代码与权重(GitHub):https://github.com/Zhouzone/SciEval/tree/main/projects/NLOS
- 🧩 LoRA 精修(进行中):在论文 training-free 重建上增补轻量 LoRA 的方案与脚手架见 GitHub 的
projects/NLOS/lora_finetune/(img2img + LoRA,多底模出多风格)。 - 📦 数据与结果(本仓库):因含约 6.5 万个小图片文件,为避免逐个上传触发限流,全部数据打包为单个档案
nlos.tar.gz(约 1.9 GB),解压后得到nlos/文件夹。 - 🟦 投影仪遮黑版
nlos_blocked.tar.gz(约 1.1 GB):在采集图固定区域覆盖黑块遮挡投影仪,解压得到nlos_blocked/(含train_blocked/test_blocked/subtracted_*_blocked/background_blocked.jpg)。对应脚本train_cls_blocked.py/test_cls_blocked.py。
下载与解压
本仓库提供两个档案,按需下载:
| 档案 | 大小 | 解压得到 | 对应脚本 |
|---|---|---|---|
nlos.tar.gz |
~1.9 GB | nlos/ |
train_cls.py / test_cls.py |
nlos_blocked.tar.gz |
~1.1 GB | nlos_blocked/ |
train_cls_blocked.py / test_cls_blocked.py |
from huggingface_hub import hf_hub_download
# 原始版(如需遮黑版改 filename="nlos_blocked.tar.gz")
path = hf_hub_download(
repo_id="Cayenne226/SciEval-NLOS",
repo_type="dataset",
filename="nlos.tar.gz",
local_dir="./",
)
tar -xzf nlos.tar.gz # Linux / macOS / Windows 10+ 自带 tar
tar -xzf nlos_blocked.tar.gz # 遮黑版
解压后将各子文件夹路径填入对应脚本顶部变量即可。
目录结构(解压后 nlos/)
| 路径 | 含义 |
|---|---|
img_train/ img_test/ |
原始 STL-10 训练集 / 测试集图像(复原目标 ground truth) |
train/ test/ |
布置 NLOS 场景后实际采集到的投影图像 |
subtracted_train/ subtracted_test/ |
将 train / test 减去 background.jpg 后的数据(消除静态背景) |
result_train100/ |
训练集复原结果(1/ 为 GT/input/output 三联对比图,2/ 为纯复原图) |
result_test100/ |
测试集复原结果(结构同上) |
background.jpg |
采集场景的纯背景图,用于背景扣除 |
README.md |
本说明文档(档案内同样附带一份) |
数据共 10 个类别:airplane, bird, car, cat, deer, dog, horse, monkey, ship, truck,各数据集内部按类别分子文件夹。
投影仪遮黑版(解压后 nlos_blocked/)
在采集图固定区域(投影仪所在位置,640×480 坐标下 (414,142)-(522,184))覆盖黑色矩形,遮挡投影仪本体与镜头亮光这一静态干扰,其余与原数据一致。
| 路径 | 含义 |
|---|---|
train_blocked/ test_blocked/ |
NLOS 采集投影图(投影仪已遮黑),各 5000 / 8000 张 |
subtracted_train_blocked/ subtracted_test_blocked/ |
背景扣除图(投影仪已遮黑) |
result_train100_blocked/ result_test100_blocked/ |
用遮黑数据训练/测试得到的复原结果(1/ 三联对比图,2/ 纯复原图) |
background_blocked.jpg |
遮黑后的背景图 |
README.md |
遮黑版说明(档案内附带) |
注:
nlos_blocked/的输入侧为遮黑数据;目标图沿用原版nlos/中的img_train/img_test(STL-10 原图,本身无投影仪),训练/测试时两者配合使用。
处理流程
原始图 (img_*) ──布置 NLOS 场景采集──▶ 投影图 (train/test)
│
减去 background.jpg
▼
背景扣除图 (subtracted_*)
│
输入网络训练/推理
▼
复原的语义轮廓图 (result_*)
网络以 subtracted_train(背景扣除图)为输入、img_train(原始图)为目标,
采用「回归(MSE 复原)+ 分类(CrossEntropy)」多任务训练,复原出带 label 信息的轮廓图。
训练与推理代码、网络结构(URnet7)及权重 train100.pth 见上方 GitHub 链接。
- Downloads last month
- 9