Upload Ti10Mo6Cu LPBF VQA dataset

README.md

@@ -0,0 +1,3 @@
+# Ti10Mo6Cu LPBF VQA dataset
+
+Each Parquet row has the seven fields `query`, `image`, `annot`, `reasoning`, `cate`, `task`, and `metadata`. `image` is a Hugging Face-compatible `{bytes, path}` image struct containing an RGB JPEG. The source spreadsheet supplies measured condition-level averages; all images for a power/speed condition are in the same split.

dataset_report.json

@@ -0,0 +1,14 @@
+{
+  "total_records": 287,
+  "conditions": 48,
+  "test_conditions": 10,
+  "skipped": [],
+  "splits": {
+    "train": 227,
+    "test": 60
+  },
+  "views": {
+    "cross_section": 143,
+    "top_view": 144
+  }
+}

test.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf763b49f694bc9c66c3202bf45566f7982a6b1aaa0ebed1365f8e1a352d62dd
+size 67041998

train.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:057b4cf80b1cd13640df88d3d718fd65074b6d8dbe427e7acf7321a5050cc006
+size 251111503