Upload DED thermal VQA dataset

README.md

@@ -0,0 +1,7 @@
+# DED Thermal VQA Dataset
+
+This dataset contains embedded thermal images from laser directed energy deposition.
+
+Fields: `query`, `image`, `annot`, `reasoning`, `cate`, `task`, `metadata`.
+
+The source images are 16-bit grayscale TIFF files. The embedded images are normalized JPEG visualizations, while raw intensity statistics are retained in `metadata`.

dataset_report.json

@@ -0,0 +1,7 @@
+{
+  "total_records": 203602,
+  "train_records": 135545,
+  "test_records": 68057,
+  "skipped_records": 2,
+  "nested_archives": 9
+}

test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:692977bc7ddbfb341a3c9a14a6441acc55dd52ce137b63bcc9d4b5c9585bbe7e
+size 833085699

train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:751275e15715fcaffa4d9af885b64104fdcd52c702b2a00e334307b0a904d0e1
+size 1623357588