| { |
| "id": "single-software-execution--cli--blender--task-39--ubuntu", |
| "snapshot": "Blender-4.2.3", |
| "instruction": "Scan → Retopo → Unwrap → Bake → Render (the flagship chain) `/home/user/Desktop/scene.blend` contains a single high-poly mesh object named `Scan` (≥ 800 vertices) imported from `/home/user/Desktop/scan.obj`. Launch Blender by running `blender` from a terminal. For this workflow, use `/opt/blender/blender` for Blender command-line or background runs when working on `scene.blend`.\n\nThe scan is bumpy — it carries surface detail you will need to preserve on the normal map. There is no camera, no light, no retopology mesh, no UV layer, and no material. Produce the full five-stage pipeline and save the result as `/home/user/Desktop/answer.blend` with two PNGs directly on the Desktop: ## Required stages ### 1. Retopo Create a new mesh object named exactly `Retopo` that retopologises the `Scan`. The `Retopo` mesh must: - have between 300 and 1200 vertices (inclusive) - approximately match the silhouette of `Scan` (it does not need to be pixel-perfect — a low-poly proxy such as a decimated copy, a shrink-wrapped icosphere, or a quad remesh is acceptable). ### 2. Unwrap The `Retopo` mesh must have an **active UV layer** whose every UV coordinate lies within the unit square `[0, 1]^2` (with a tiny float tolerance). Any unwrap technique is fine: Smart UV Project, Cube Projection, or a manual unwrap followed by Pack Islands and a bbox-normalising scale. ### 3. Normal bake (tangent-space) Bake a tangent-space normal map from `Scan` onto `Retopo`'s UV layout and save it as `normal_bake.png` at **512×512**. The bake must satisfy, when the PNG is read as non-color data: - mean B channel ≥ 0.80 (i.e. most of the surface is pointing outward along the tangent +Z axis, as expected for tangent-space normals), - mean R channel in [0.40, 0.60] and mean G channel in [0.40, 0.60] (centred around neutral 0.5), - std R ≥ 0.02 and std G ≥ 0.02 (actual per-texel surface variation was captured — a flat 0.5-grey image does **not** pass). Use Cycles for the bake, set `render.bake.use_selected_to_active=True`, cage 0.01–0.05 BU, and ensure the target image is assigned as the active image texture on `Retopo`'s material. ### 4. Scene setup Add a camera and at least one light such that `Retopo` is framed and lit. Material on `Retopo` must sample `normal_bake.png` through a Normal Map node feeding a Principled BSDF. ### 5. Hero render Render a final image at **512×512** named `hero.png`. Mean luma (Rec.709-ish weighting 0.2126·R + 0.7152·G + 0.0722·B) over the entire frame must fall inside the band [0.10, 0.70] — i.e. the render is neither pitch-black nor blown-out. ## Requirements 1. `/home/user/Desktop/answer.blend` exists and opens. 2. Mesh object named `Retopo` is present. 3. `Retopo` has 300–1200 vertices. 4. `Retopo` has an active UV layer. 5. Every UV coordinate on that layer lies in [0, 1]². 6. `normal_bake.png` exists at 512×512. 7. `normal_bake.png` satisfies the mean-B ≥ 0.80, mean-RG ∈ [0.40, 0.60], std-RG ≥ 0.02 constraints. 8. `hero.png` exists at 512×512. 9. Mean luma of `hero.png` is in [0.10, 0.70]. 10. The starting `Scan` mesh need not be retained — only `Retopo`'s state is checked.", |
| "source": "https://docs.blender.org/manual/en/4.1/; https://docs.blender.org/api/4.1/; https://developer.blender.org/docs/release_notes/4.1/; https://www.loc.gov/preservation/digital/formats/fdd/fdd000507.shtml", |
| "config": [ |
| { |
| "type": "upload_file", |
| "parameters": { |
| "files": [ |
| { |
| "local_path": "task-39/init_file/scan.obj", |
| "path": "/home/user/Desktop/scan.obj" |
| } |
| ] |
| } |
| }, |
| { |
| "type": "upload_file", |
| "parameters": { |
| "files": [ |
| { |
| "local_path": "task-39/init_file/scene.blend", |
| "path": "/home/user/Desktop/scene.blend" |
| } |
| ] |
| } |
| } |
| ], |
| "trajectory": "trajectories/", |
| "related_apps": [ |
| "blender" |
| ], |
| "artifact_paths": [ |
| "/home/user/Desktop/answer.blend", |
| "/home/user/Desktop/hero.png", |
| "/home/user/Desktop/normal_bake.png" |
| ], |
| "evaluator": { |
| "postconfig": [ |
| { |
| "type": "upload_file", |
| "parameters": { |
| "files": [ |
| { |
| "local_path": "task-39/eval.py", |
| "path": "/home/user/Desktop/eval.py" |
| } |
| ] |
| } |
| } |
| ], |
| "func": "exact_match", |
| "result": { |
| "type": "vm_command_line", |
| "command": "python3 /home/user/Desktop/eval.py", |
| "shell": "true" |
| }, |
| "expected": { |
| "type": "rule", |
| "rules": { |
| "expected": "True\n" |
| } |
| } |
| }, |
| "proxy": false, |
| "fixed_ip": false, |
| "possibility_of_env_change": "low" |
| } |
|
|
