Limit the sizes of some of the nightly examples (#4545)

They are still not exactlty _small_, but a lot _smaller_, and since they
all stream, it should be pretty fine anyways

```
detect_and_track_objects.rrd (55.3 MiB)
nuscenes.rrd (90.0 MiB)
objectron.rrd (81.1 MiB)
open_photogrammetry_format.rrd (67.6 MiB)
raw_mesh.rrd (116 kiB)
rgbd.rrd (36.8 MiB)
segment_anything_model.rrd (8.5 MiB)
```

This helped in figuring out what to focus on:
* #4542

### Checklist
* [x] I have read and agree to [Contributor
Guide](https://github.com/rerun-io/rerun/blob/main/CONTRIBUTING.md) and
the [Code of
Conduct](https://github.com/rerun-io/rerun/blob/main/CODE_OF_CONDUCT.md)
* [x] I've included a screenshot or gif (if applicable)
* [x] I have tested the web demo (if applicable):
* Using newly built examples:
[app.rerun.io](https://app.rerun.io/pr/4545/index.html)
* Using examples from latest `main` build:
[app.rerun.io](https://app.rerun.io/pr/4545/index.html?manifest_url=https://app.rerun.io/version/main/examples_manifest.json)
* Using full set of examples from `nightly` build:
[app.rerun.io](https://app.rerun.io/pr/4545/index.html?manifest_url=https://app.rerun.io/version/nightly/examples_manifest.json)
* [x] The PR title and labels are set such as to maximize their
usefulness for the next release's CHANGELOG

- [PR Build Summary](https://build.rerun.io/pr/4545)
- [Docs
preview](https://rerun.io/preview/4aa4c24e403c16b455d8be8b02b98d5c8164fa53/docs)
<!--DOCS-PREVIEW-->
- [Examples
preview](https://rerun.io/preview/4aa4c24e403c16b455d8be8b02b98d5c8164fa53/examples)
<!--EXAMPLES-PREVIEW-->
- [Recent benchmark results](https://build.rerun.io/graphs/crates.html)
- [Wasm size tracking](https://build.rerun.io/graphs/sizes.html)

Author: emilk

examples/python/nuscenes/README.md

@@ -6,6 +6,7 @@ description: "Visualize the nuScenes dataset including lidar, radar, images, and
 thumbnail: https://static.rerun.io/nuscenes/64a50a9d67cbb69ae872551989ee807b195f6b5d/480w.png
 thumbnail_dimensions: [480, 282]
 channel: nightly
+build_args: ["--seconds=5"]
 ---
 
 <picture>

examples/python/nuscenes/main.py

@@ -47,7 +47,7 @@ def ensure_scene_available(root_dir: pathlib.Path, dataset_version: str, scene_n
         raise ValueError(f"{scene_name=} not found in dataset")
 
 
-def log_nuscenes(root_dir: pathlib.Path, dataset_version: str, scene_name: str) -> None:
+def log_nuscenes(root_dir: pathlib.Path, dataset_version: str, scene_name: str, max_time_sec: float) -> None:
     """Log nuScenes scene."""
     nusc = nuscenes.NuScenes(version=dataset_version, dataroot=root_dir, verbose=True)
 
@@ -72,20 +72,26 @@ def log_nuscenes(root_dir: pathlib.Path, dataset_version: str, scene_name: str)
         elif sample_data["sensor_modality"] == "camera":
             first_camera_tokens.append(sample_data_token)
 
-    log_lidar_and_ego_pose(first_lidar_token, nusc)
-    log_cameras(first_camera_tokens, nusc)
-    log_radars(first_radar_tokens, nusc)
-    log_annotations(first_sample_token, nusc)
+    first_timestamp_us = nusc.get("sample_data", first_lidar_token)["timestamp"]
+    max_timestamp_us = first_timestamp_us + 1e6 * max_time_sec
 
+    log_lidar_and_ego_pose(first_lidar_token, nusc, max_timestamp_us)
+    log_cameras(first_camera_tokens, nusc, max_timestamp_us)
+    log_radars(first_radar_tokens, nusc, max_timestamp_us)
+    log_annotations(first_sample_token, nusc, max_timestamp_us)
 
-def log_lidar_and_ego_pose(first_lidar_token: str, nusc: nuscenes.NuScenes) -> None:
+
+def log_lidar_and_ego_pose(first_lidar_token: str, nusc: nuscenes.NuScenes, max_timestamp_us: float) -> None:
     """Log lidar data and vehicle pose."""
     current_lidar_token = first_lidar_token
 
     while current_lidar_token != "":
         sample_data = nusc.get("sample_data", current_lidar_token)
         sensor_name = sample_data["channel"]
 
+        if max_timestamp_us < sample_data["timestamp"]:
+            break
+
         # timestamps are in microseconds
         rr.set_time_seconds("timestamp", sample_data["timestamp"] * 1e-6)
 
@@ -109,25 +115,29 @@ def log_lidar_and_ego_pose(first_lidar_token: str, nusc: nuscenes.NuScenes) -> N
         rr.log(f"world/ego_vehicle/{sensor_name}", rr.Points3D(points, colors=point_colors))
 
 
-def log_cameras(first_camera_tokens: list[str], nusc: nuscenes.NuScenes) -> None:
+def log_cameras(first_camera_tokens: list[str], nusc: nuscenes.NuScenes, max_timestamp_us: float) -> None:
     """Log camera data."""
     for first_camera_token in first_camera_tokens:
         current_camera_token = first_camera_token
         while current_camera_token != "":
             sample_data = nusc.get("sample_data", current_camera_token)
+            if max_timestamp_us < sample_data["timestamp"]:
+                break
             sensor_name = sample_data["channel"]
             rr.set_time_seconds("timestamp", sample_data["timestamp"] * 1e-6)
             data_file_path = nusc.dataroot / sample_data["filename"]
             rr.log(f"world/ego_vehicle/{sensor_name}", rr.ImageEncoded(path=data_file_path))
             current_camera_token = sample_data["next"]
 
 
-def log_radars(first_radar_tokens: list[str], nusc: nuscenes.NuScenes) -> None:
+def log_radars(first_radar_tokens: list[str], nusc: nuscenes.NuScenes, max_timestamp_us: float) -> None:
     """Log radar data."""
     for first_radar_token in first_radar_tokens:
         current_camera_token = first_radar_token
         while current_camera_token != "":
             sample_data = nusc.get("sample_data", current_camera_token)
+            if max_timestamp_us < sample_data["timestamp"]:
+                break
             sensor_name = sample_data["channel"]
             rr.set_time_seconds("timestamp", sample_data["timestamp"] * 1e-6)
             data_file_path = nusc.dataroot / sample_data["filename"]
@@ -139,14 +149,16 @@ def log_radars(first_radar_tokens: list[str], nusc: nuscenes.NuScenes) -> None:
             current_camera_token = sample_data["next"]
 
 
-def log_annotations(first_sample_token: str, nusc: nuscenes.NuScenes) -> None:
+def log_annotations(first_sample_token: str, nusc: nuscenes.NuScenes, max_timestamp_us: float) -> None:
     """Log 3D bounding boxes."""
     label2id: dict[str, int] = {}
     current_sample_token = first_sample_token
     while current_sample_token != "":
-        sample = nusc.get("sample", current_sample_token)
-        rr.set_time_seconds("timestamp", sample["timestamp"] * 1e-6)
-        ann_tokens = sample["anns"]
+        sample_data = nusc.get("sample", current_sample_token)
+        if max_timestamp_us < sample_data["timestamp"]:
+            break
+        rr.set_time_seconds("timestamp", sample_data["timestamp"] * 1e-6)
+        ann_tokens = sample_data["anns"]
         sizes = []
         centers = []
         rotations = []
@@ -164,7 +176,7 @@ def log_annotations(first_sample_token: str, nusc: nuscenes.NuScenes) -> None:
             class_ids.append(label2id[ann["category_name"]])
 
         rr.log("world/anns", rr.Boxes3D(sizes=sizes, centers=centers, rotations=rotations, class_ids=class_ids))
-        current_sample_token = sample["next"]
+        current_sample_token = sample_data["next"]
 
     # skipping for now since labels take too much space in 3D view (see https://github.com/rerun-io/rerun/issues/4451)
     # annotation_context = [(i, label) for label, i in label2id.items()]
@@ -213,13 +225,16 @@ def main() -> None:
         help="Scene name to visualize (typically of form 'scene-xxxx')",
     )
     parser.add_argument("--dataset-version", type=str, default="v1.0-mini", help="Scene id to visualize")
+    parser.add_argument(
+        "--seconds", type=float, default=float("inf"), help="If specified, limits the number of seconds logged"
+    )
     rr.script_add_args(parser)
     args = parser.parse_args()
 
     ensure_scene_available(args.root_dir, args.dataset_version, args.scene_name)
 
     rr.script_setup(args, "rerun_example_nuscenes")
-    log_nuscenes(args.root_dir, args.dataset_version, args.scene_name)
+    log_nuscenes(args.root_dir, args.dataset_version, args.scene_name, max_time_sec=args.seconds)
 
     rr.script_teardown(args)
 

examples/python/objectron/README.md

@@ -6,6 +6,7 @@ tags: [2D, 3D, object-detection, pinhole-camera]
 thumbnail: https://static.rerun.io/objectron/8ea3a37e6b4af2e06f8e2ea5e70c1951af67fea8/480w.png
 thumbnail_dimensions: [480, 268]
 channel: nightly
+build_args: ["--frames=150"]
 ---
 
 <picture>

examples/python/open_photogrammetry_format/README.md

@@ -5,6 +5,7 @@ tags: [2d, 3d, camera, photogrammetry]
 thumbnail: https://static.rerun.io/open_photogrammetry_format/603d5605f9670889bc8bce3365f16b831fce1eb1/480w.png
 thumbnail_dimensions: [480, 310]
 channel: nightly
+build_args: ["--jpeg-quality=50"]
 ---
 
 <picture>

examples/python/open_photogrammetry_format/main.py

@@ -21,6 +21,7 @@
 import requests
 import rerun as rr
 import tqdm
+from PIL import Image
 from pyopf.io import load
 from pyopf.resolve import resolve
 
@@ -109,10 +110,10 @@ def from_dataset(cls, dataset: str, log_as_frames: bool = True) -> OPFProject:
 
     def log_point_cloud(self) -> None:
         """Log the project's point cloud."""
-        pcl = self.project.point_cloud_objs[0]
-        rr.log("world/pcl", rr.Points3D(pcl.nodes[0].position, colors=pcl.nodes[0].color), timeless=True)
+        points = self.project.point_cloud_objs[0].nodes[0]
+        rr.log("world/points", rr.Points3D(points.position, colors=points.color), timeless=True)
 
-    def log_calibrated_cameras(self) -> None:
+    def log_calibrated_cameras(self, jpeg_quality: int | None) -> None:
         """
         Log the project's calibrated cameras as individual frames.
 
@@ -180,7 +181,12 @@ def log_calibrated_cameras(self) -> None:
                     camera_xyz=rr.ViewCoordinates.RUB,
                 ),
             )
-            rr.log(entity + "/image/rgb", rr.ImageEncoded(path=self.path.parent / camera.uri))
+
+            if jpeg_quality is not None:
+                with Image.open(self.path.parent / camera.uri) as img:
+                    rr.log(entity + "/image/rgb", rr.Image(np.array(img)).compress(jpeg_quality=jpeg_quality))
+            else:
+                rr.log(entity + "/image/rgb", rr.ImageEncoded(path=self.path.parent / camera.uri))
 
 
 def main() -> None:
@@ -199,6 +205,12 @@ def main() -> None:
         action="store_true",
         help="Log all cameras globally instead of as individual frames in the timeline.",
     )
+    parser.add_argument(
+        "--jpeg-quality",
+        type=int,
+        default=None,
+        help="If specified, compress the camera images with the given JPEG quality.",
+    )
 
     rr.script_add_args(parser)
 
@@ -213,7 +225,7 @@ def main() -> None:
     rr.script_setup(args, "rerun_example_open_photogrammetry_format")
     rr.log("world", rr.ViewCoordinates.RIGHT_HAND_Z_UP, timeless=True)
     project.log_point_cloud()
-    project.log_calibrated_cameras()
+    project.log_calibrated_cameras(jpeg_quality=args.jpeg_quality)
     rr.script_teardown(args)
 
 

examples/python/open_photogrammetry_format/requirements.txt

@@ -1,4 +1,5 @@
 numpy
+pillow
 pyopf; python_version >= '3.10'  # silence error if this requirements.txt gets pulled in a <3.10 venv
 requests
 rerun-sdk

examples/python/rgbd/README.md

@@ -5,6 +5,7 @@ tags: [2D, 3D, depth, nyud, pinhole-camera]
 thumbnail: https://static.rerun.io/rgbd/4109d29ed52fa0a8f980fcdd0e9673360c76879f/480w.png
 thumbnail_dimensions: [480, 254]
 channel: nightly
+build_args: ["--frames=300"]
 ---
 
 <picture>

examples/python/rgbd/main.py

@@ -8,6 +8,7 @@
 
 import argparse
 import os
+import sys
 import zipfile
 from datetime import datetime
 from pathlib import Path
@@ -50,7 +51,7 @@ def read_depth_image(buf: bytes) -> npt.NDArray[Any]:
     return img
 
 
-def log_nyud_data(recording_path: Path, subset_idx: int = 0) -> None:
+def log_nyud_data(recording_path: Path, subset_idx: int, frames: int) -> None:
     rr.log("world", rr.ViewCoordinates.RIGHT_HAND_Y_DOWN, timeless=True)
 
     with zipfile.ZipFile(recording_path, "r") as archive:
@@ -67,6 +68,9 @@ def log_nyud_data(recording_path: Path, subset_idx: int = 0) -> None:
         files_with_timestamps = [(parse_timestamp(f.filename), f) for f in subset]
         files_with_timestamps.sort(key=lambda t: t[0])
 
+        if len(files_with_timestamps) > frames:
+            files_with_timestamps = files_with_timestamps[:frames]
+
         for time, f in files_with_timestamps:
             rr.set_time_seconds("time", time.timestamp())
 
@@ -140,7 +144,7 @@ def download_progress(url: str, dst: Path) -> None:
             bar.update(size)
 
 
-if __name__ == "__main__":
+def main() -> None:
     parser = argparse.ArgumentParser(description="Logs rich data using the Rerun SDK.")
     parser.add_argument(
         "--recording",
@@ -150,6 +154,9 @@ def download_progress(url: str, dst: Path) -> None:
         help="Name of the NYU Depth Dataset V2 recording",
     )
     parser.add_argument("--subset-idx", type=int, default=0, help="The index of the subset of the recording to use.")
+    parser.add_argument(
+        "--frames", type=int, default=sys.maxsize, help="If specified, limits the number of frames logged"
+    )
     rr.script_add_args(parser)
     args = parser.parse_args()
 
@@ -159,6 +166,11 @@ def download_progress(url: str, dst: Path) -> None:
     log_nyud_data(
         recording_path=recording_path,
         subset_idx=args.subset_idx,
+        frames=args.frames,
     )
 
     rr.script_teardown(args)
+
+
+if __name__ == "__main__":
+    main()

examples/rust/objectron/README.md

@@ -4,6 +4,7 @@ python: https://github.com/rerun-io/rerun/tree/latest/examples/python/objectron/
 rust: https://github.com/rerun-io/rerun/tree/latest/examples/rust/objectron/src/main.rs
 tags: [2D, 3D, object-detection, pinhole-camera]
 thumbnail: https://static.rerun.io/objectron/8ea3a37e6b4af2e06f8e2ea5e70c1951af67fea8/480w.png
+build_args: ["--frames=100"]
 ---
 
 <picture>