# Step 4: 2D Annotation + QA Now we annotate the selected batch. This step covers: 1. Setting up a consistent annotation schema (KITTI classes) 2. Annotating 2D detections on the **left camera slice** 3. QA checks before moving to 3D annotation > **Time commitment:** Plan 1-2 minutes per scene for careful annotation. Start with 10-20 scenes to get the workflow, then continue or use the fast-forward option. ## Define Your Schema (KITTI Classes) Before labeling, define the rules. This prevents class drift and maintains consistency. We use the standard KITTI classes for autonomous driving. ## Annotate 2D Detections in the App **This is the real labeling step.** Open the App and annotate the left camera images. ### Setup (one time) 1. Launch the App with your batch 2. Click a sample to open the modal 3. **Select the \`\`left\`\` slice** from the slice dropdown 4. Click the **Annotate** tab (pencil icon) 5. Click **Schema** -> **New Field** -> name it `human_detections` 6. Set type to **Detections** and add the KITTI classes above ### For each scene 1. Ensure you’re on the **left** slice 2. Review the image 3. Click **Detection** button (square icon) 4. Draw boxes around all vehicles, pedestrians, cyclists 5. Assign the correct KITTI class 6. Move to the next scene ### Labeling Guidelines - **Car**: Sedans, SUVs, hatchbacks - **Van**: Minivans, cargo vans - **Truck**: Pickup trucks, semi-trucks - **Pedestrian**: Standing or walking people - **Person_sitting**: Seated people (benches, ground) - **Cyclist**: Person on bicycle - **Tram**: Streetcars, light rail - **Misc**: Ambiguous or other vehicles ### Stop here and annotate samples Take 15-30 minutes to label some scenes. This is the core skill. When you’re done (or want to fast-forward), continue below. --- ## Fast-Forward Option If you want to proceed without labeling everything manually, set `FAST_FORWARD = True` below. This copies `ground_truth` labels to `human_detections` to simulate completed annotation. > **Note:** In real projects, there’s no shortcut. Label quality determines model quality. ## Mark Annotated Samples **Important:** We only mark samples as “annotated” if they actually have labels. ## QA Checks Before moving to 3D annotation, verify 2D label quality. ## Summary You annotated 2D detections on the left camera slice: - Defined KITTI schema for consistency - Labeled samples in the App (or fast-forwarded) - **Only samples with actual labels** were marked as annotated - Ran QA checks: coverage, class distribution, schema compliance **Artifacts:** - `human_detections` field with 2D bounding boxes - `annotated_2d:v0` tag on left camera samples with labels - `annotated:v0` tag on all slices of annotated groups **Next:** Step 5 - 3D Annotation (cuboids on point clouds)