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/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 290, in _generate_tables
                  pa_table = paj.read_json(
                      io.BytesIO(batch), read_options=paj.ReadOptions(block_size=block_size)
                  )
                File "pyarrow/_json.pyx", line 342, in pyarrow._json.read_json
                File "pyarrow/error.pxi", line 155, in pyarrow.lib.pyarrow_internal_check_status
                File "pyarrow/error.pxi", line 92, in pyarrow.lib.check_status
                  raise convert_status(status)
              pyarrow.lib.ArrowInvalid: JSON parse error: Column() changed from object to string in row 0
              
              During handling of the above exception, another exception occurred:
              
              Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 101, in _split_generators
                  pa_table = next(iter(self._generate_tables(**splits[0].gen_kwargs, allow_full_read=False)))[1]
                             ~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 304, in _generate_tables
                  batch = json_encode_fields_in_json_lines(original_batch, json_field_paths)
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 111, in json_encode_fields_in_json_lines
                  examples = [ujson_loads(line) for line in original_batch.splitlines()]
                              ~~~~~~~~~~~^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/utils/json.py", line 20, in ujson_loads
                  return pd.io.json.ujson_loads(*args, **kwargs)
                         ~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^
              ValueError: Expected object or value
              
              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 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/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.

YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

# 数据集说明（AdcSR 微调/蒸馏用 · CSIG 赛题二轻量Diffusion 超分）

目标域: 4K 城市建筑 / 夜景建筑 / 高楼 / 商铺街道 / 花草 / 居民楼 / 城市远景 / 其他建筑训练口径: 官方 AdcSR 4x 链, 输入 LR 128x128 -> 输出 HR 512x512 (P1: 512窗->bicubic128->模型->512->融合) 本数据集仅供训练/蒸馏; 推理与官方 3 对 val 见工程根 README/NOTES。

1. 数据源与角色

源	类型	角色
NKUSR8K (8K 植物/建筑/街道, 无夜景)	纯 HR	GT patch 池
4KLSDB (train_x4 parquet, 原生 4K)	纯 HR(带 caption)	GT patch 池(建筑/风景/花草, 夜景适量)
4KLSDB 原图 URL 定向抓取	纯 HR(1~6K)	稀疏类补量(夜建/商场/店铺/居民/高楼/城景/其他建筑)
LIU4K v2 (Building/Street/Mountain)	纯 HR(4K+ PNG)	GT patch 池(建筑/街道/风景)
RealSR V3 (真实 LR/HR 对)	真实退化对	real_pairs 训练/验证(退化域适配)

2. 目录布局(data/, 整理后)

clean_all/<SRC>/: 清洗后纯 HR 原图 jpg(最长边 4096, q92); ∈ {NKUSR8K, LIU4K, 4KLSDB_b1, 4KLSDB_b2, 4KLSDB_url}
real_pairs/RealSR/{train,test}_{hr,lr}/: RealSR V3 真实对(HR~~1400x800, LR~~350x200)
patches/<category>/: 离线随机裁 512x512 HR patch(仅训练源, 文件名 <source>_<stem>__<k>.jpg)
patches_val/: 源级隔离的 val patch(训练零重叠)
manifest_train.json / manifest_val.json: 训练/验证清单(synthetic_hr + real_pairs)
logs/cat_pool_final.json: 最终 8 类池报表(每项 {path}); 同目录 *_clip.json 为各类别 CLIP 打标
review/pool_all/*.jpg: 每类 20 张人工复核拼图
split_val_sources.json: 留作 val 的源图清单(按源图分组, 防同源泄露)

3. 清洗规则(全部落脚本 tools/)

短边 <960 剔除; 宽高比超出 [0.35, 3.0] 剔除
灰度拉普拉斯方差(512 缩放下)<50 判模糊剔除
纯色/全黑(std<10 或非零像素<0.5%)、过曝(均值>245 或 >30% 像素>250)、低熵(<3.5)剔除
aHash/dHash 去重(clean_images 内 aHash; 全池最终 pHash/dHash 见 tools/dedup_phash.py)
CLIP(openai/clip-vit-base-patch32) 8 类分类 + 剔除 people/indoor/other + night-gate
JPEG 重压体积 <20KB 阈值: 由 min_short>=960+entropy>=3.5 覆盖(可加 --min_recompressed)

4. 8 类配额(清洗后 HR 原图目标)

night_building(夜晚风光建筑) / skyscraper(大厦高楼) / storefront(店铺门牌) / mall(商场外拍) / plant(花草树木) / residential(居民楼) / city_view(城市远景) / other_building(其他建筑) 目标: 每类 400-900, 总池 5000-7000(达标检查见 logs/SUMMARY.txt / README 底部计数)

5. 训练用法

HR patch 来自 data/patches; real pairs 来自 data/real_pairs(与 HR patch 混训, --data_mode mix)
在线增强(训练时): Real-ESRGAN 风格退化(LR128 由 512 HR patch 生成) + 翻转/旋转/色彩抖动 (训练脚本 src/train_lora.py / train_stage2.py; 具体退化参数与 RealPairDataset 对齐)
代理评测: RealSR test 99 对 + val 源 patch; 官方 3 对仅终检

6. 复现数据管线(工具链)

clean_images.py -> (tag_categories/clip_filter.py) -> merge_reports.py -> dedup_phash.py -> split_val.py(源级) -> build_patches2.py -> build_manifest.py 4KLSDB 抽取: klsdb_index.py -> klsdb_extract.py(rank/extract); URL 补量: fetch_klsdb_urls.py LIU4K: gdown 分卷 + NVIDIA 7-Zip 解压 -> clean_images -> convert_png_to_jpg.py

7. 版本与计数

最后更新: 见 logs/cat_pool_final.json summary 与 data/SUMMARY.txt(由 finalize 生成)
git 提交记录见工程根 NOTES.md

Downloads last month: 47

Total file size:

27.7 GB