- Rating
- 3
- Confidence
- 3
- Informativeness
- 2
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
ALF uses quantized detection boxes as a shared communication interface across heterogeneous agents. It converts received boxes into pseudo-BEV maps and combines them with ego features to synthesize features for intermediate fusion. This allows collaboration with unseen LiDAR specifications or encoder configurations without further adaptation.
Critical strengths
Reducing the adaptation cost of integrating new agents is practically relevant. The approach is straightforward and communication-efficient, requiring only 120 bytes of box data per agent per frame on V2X-Real. In the zero-shot evaluation, ALF achieves an average [email protected] of 0.311, compared with 0.229 for the strongest baseline. Comparisons with late fusion using the same boxes also support the benefit of feature synthesis.
Critical weaknesses
-
The zero-shot comparison may be affected by how the baselines are implemented. Appendix B.5 uses an untrained, orthogonally initialized convolution to match feature channels when introducing a new encoder. This mapping may itself disrupt feature semantics, so the performance drop cannot be attributed solely to configuration changes. An identity-mapping control for configurations with matching channel dimensions would help isolate this effect. Table 8 also needs clarification: ALF’s results vary with AUX2 even though evaluation uses a fixed EGO+AUX1 pair and ALF undergoes no adaptation.
-
The information lost through box compression and the limits of feature synthesis are insufficiently examined. Each message contains only planar position, length, width, yaw, and confidence, omitting class labels, vertical position, and height despite the multi-class 3D detection task. How reliably can these missing attributes be recovered when the object is occluded from the ego vehicle? Box dropout and noise experiments do not fully address this question. Results by class and occlusion level, together with comparisons against messages retaining class and height information, would be useful.
-
Evidence for open-world deployment remains limited. Unseen configurations mainly involve PointPillar and SECOND variants and 64/128-beam LiDARs, while the ego configuration remains fixed. A shared message format ensures interface compatibility but does not remove differences in detector confidence or error distributions. Evaluation with detectors trained on different datasets or substantially different architectures would strengthen the generalization claim. The paper should also report inference time, memory usage, and how synthesis costs scale with the number of collaborators.
-
Relevant prior work is missing. GT-Space: Enhancing Heterogeneous Collaborative Perception with Ground Truth Feature Space also addresses feature compatibility and the integration of new agents. It constructs a shared feature space from ground truth and trains adapters for new configurations. The authors should explain how ALF’s ego-specific feature space differs and compare detection accuracy, communication cost, and adaptation cost under matched data and training conditions.
Questions for authors
see Weakness.
8 comments
Response to Reviewer wjK1 (2/6)
W2: Effects of adding class, vertical position, and height to box messages [1/2]
To evaluate the effect of retaining class, vertical position, and height in the box message, we compare 6-field ALF and 9-field ALF. We measure class-wise detection AP and recall under limited Ego visibility in this post, and evaluate the accuracy of the final box attributes in [2/2].
Experimental setup
The six-field message contains planar position (x, y), length, width, yaw, and confidence. The nine-field message adds the auxiliary detector's predicted class, vertical position z, and box height. We train the two versions of the Ego-compatible Feature Synthesizer (EFS) under matched conditions with the receiving model fixed. We evaluate both on V2X-Real using PP6, PP8, SD2, and SD3 as auxiliary configurations.
Detection performance by class
Each cell reports [email protected] / 0.7. mAP is the mean over the three classes. Bold marks the higher value for each threshold using the unrounded scores.
| Metric | ALF, 6 fields | ALF, 9 fields |
|---|---|---|
| mAP | 0.480 / 0.310 | 0.480 / 0.309 |
| Vehicle AP | 0.735 / 0.540 | 0.741 / 0.549 |
| Pedestrian AP | 0.182 / 0.015 | 0.181 / 0.015 |
| Truck AP | 0.523 / 0.374 | 0.517 / 0.362 |
At IoU 0.7, the nine-field message improves Vehicle AP from 0.540 to 0.549. Truck AP decreases from 0.374 to 0.362, both Pedestrian AP values round to 0.015, and mAP changes from 0.310 to 0.309.
Recall by class and Ego LiDAR visibility
We count Ego LiDAR returns inside each ground-truth 3D box. Zero denotes boxes with no returns. Sparse denotes boxes with at least one return and a count at or below a threshold that accounts for object projected area, class, and distance from Ego. We derive the thresholds from the lower quartile of positive point density on the validation split and hold them fixed for testing. We use these groups as a LiDAR-based proxy for Ego visibility.
Each cell reports class-aware [email protected] / 0.7. GT counts ground-truth objects once across the four AUX evaluations.
| Ego visibility | Class | GT | 6 fields | 9 fields |
|---|---|---|---|---|
| Zero | Vehicle | 2,360 | 0.741 / 0.506 | 0.736 / 0.519 |
| Zero | Pedestrian | 5,791 | 0.067 / 0.007 | 0.067 / 0.007 |
| Zero | Truck | 379 | 0.117 / 0.020 | 0.113 / 0.026 |
| Sparse | Vehicle | 4,268 | 0.821 / 0.573 | 0.820 / 0.578 |
| Sparse | Pedestrian | 10,085 | 0.281 / 0.073 | 0.282 / 0.072 |
| Sparse | Truck | 1,425 | 0.526 / 0.441 | 0.521 / 0.440 |
Across the four auxiliary configurations, mean Vehicle [email protected] increases from 0.506 to 0.519 in Zero and from 0.573 to 0.578 in Sparse. Pedestrian and Truck recall changes vary across auxiliary configurations. We will add the class-wise AP and visibility-stratified recall results to the revised manuscript.
Response to Reviewer wjK1 (3/6)
W2: Effect of transmitting class, vertical position, and height [2/2]
Class, vertical position, and height accuracy in the final detections
We match final detections one-to-one to ground truth at IoU ≥ 0.5 using box geometry without requiring class agreement. Class accuracy is the fraction of matched detections with the correct class, computed separately for each model. For vertical position () and height, we compute mean absolute error (MAE, in meters) on the same ground-truth objects matched by both models.
We average each metric equally across PP6, PP8, SD2, and SD3. Bold marks the higher class accuracy or lower MAE based on unrounded values.
| Ego visibility | Class | Message format | Class accuracy | z MAE (m) | Height MAE (m) |
|---|---|---|---|---|---|
| Zero | Vehicle | 6 fields | 0.997 | 0.152 | 0.177 |
| Zero | Vehicle | 9 fields | 0.996 | 0.152 | 0.172 |
| Zero | Pedestrian | 6 fields | 1.000 | 0.211 | 0.109 |
| Zero | Pedestrian | 9 fields | 1.000 | 0.202 | 0.108 |
| Zero | Truck | 6 fields | 0.365 | 0.340 | 0.346 |
| Zero | Truck | 9 fields | 0.350 | 0.371 | 0.328 |
| Sparse | Vehicle | 6 fields | 0.961 | 0.103 | 0.138 |
| Sparse | Vehicle | 9 fields | 0.960 | 0.102 | 0.136 |
| Sparse | Pedestrian | 6 fields | 1.000 | 0.104 | 0.090 |
| Sparse | Pedestrian | 9 fields | 1.000 | 0.103 | 0.090 |
| Sparse | Truck | 6 fields | 0.737 | 0.133 | 0.143 |
| Sparse | Truck | 9 fields | 0.737 | 0.132 | 0.146 |
With no Ego LiDAR returns, six-field ALF achieves class accuracy of 0.997 on matched vehicles and 0.365 on matched trucks. Both models correctly classify their matched Zero pedestrians, while Pedestrian [email protected] rounds to 0.007 for both formats. Adding the three fields lowers height MAE for all three Zero classes, while Truck MAE rises from 0.340 m to 0.371 m. We will add the per-class AP, visibility-stratified recall, and final-detection attribute analysis to the revised manuscript.
Results by AUX configuration
The table reports the individual results for PP6, PP8, SD2, and SD3 before averaging.
| AUX | Ego visibility | Class | Class acc. (6 fields) | Class acc. (9 fields) | MAE (6 fields) | MAE (9 fields) | Height MAE (6 fields) | Height MAE (9 fields) |
|---|---|---|---|---|---|---|---|---|
| PP6 | Zero | Vehicle | 0.997 | 0.994 | 0.140 | 0.140 | 0.163 | 0.157 |
| PP6 | Zero | Pedestrian | 1.000 | 1.000 | 0.248 | 0.234 | 0.103 | 0.101 |
| PP6 | Zero | Truck | 0.212 | 0.193 | 0.351 | 0.390 | 0.283 | 0.273 |
| PP6 | Sparse | Vehicle | 0.964 | 0.961 | 0.103 | 0.102 | 0.138 | 0.137 |
| PP6 | Sparse | Pedestrian | 1.000 | 1.000 | 0.104 | 0.104 | 0.089 | 0.089 |
| PP6 | Sparse | Truck | 0.741 | 0.743 | 0.126 | 0.123 | 0.134 | 0.137 |
| PP8 | Zero | Vehicle | 0.995 | 0.997 | 0.160 | 0.159 | 0.186 | 0.179 |
| PP8 | Zero | Pedestrian | 1.000 | 1.000 | 0.178 | 0.166 | 0.103 | 0.098 |
| PP8 | Zero | Truck | 0.370 | 0.289 | 0.359 | 0.375 | 0.359 | 0.352 |
| PP8 | Sparse | Vehicle | 0.957 | 0.958 | 0.105 | 0.104 | 0.142 | 0.140 |
| PP8 | Sparse | Pedestrian | 1.000 | 1.000 | 0.103 | 0.102 | 0.089 | 0.089 |
| PP8 | Sparse | Truck | 0.742 | 0.740 | 0.131 | 0.135 | 0.150 | 0.155 |
| SD2 | Zero | Vehicle | 0.998 | 0.997 | 0.152 | 0.150 | 0.173 | 0.170 |
| SD2 | Zero | Pedestrian | 1.000 | 1.000 | 0.203 | 0.196 | 0.112 | 0.114 |
| SD2 | Zero | Truck | 0.487 | 0.514 | 0.345 | 0.383 | 0.347 | 0.304 |
| SD2 | Sparse | Vehicle | 0.962 | 0.961 | 0.102 | 0.101 | 0.137 | 0.135 |
| SD2 | Sparse | Pedestrian | 1.000 | 1.000 | 0.103 | 0.102 | 0.090 | 0.090 |
| SD2 | Sparse | Truck | 0.738 | 0.740 | 0.135 | 0.135 | 0.140 | 0.147 |
| SD3 | Zero | Vehicle | 0.998 | 0.995 | 0.157 | 0.160 | 0.185 | 0.180 |
| SD3 | Zero | Pedestrian | 1.000 | 1.000 | 0.213 | 0.210 | 0.118 | 0.119 |
| SD3 | Zero | Truck | 0.392 | 0.405 | 0.305 | 0.337 | 0.394 | 0.385 |
| SD3 | Sparse | Vehicle | 0.959 | 0.960 | 0.102 | 0.101 | 0.136 | 0.134 |
| SD3 | Sparse | Pedestrian | 1.000 | 1.000 | 0.106 | 0.106 | 0.091 | 0.091 |
| SD3 | Sparse | Truck | 0.729 | 0.727 | 0.139 | 0.137 | 0.148 | 0.143 |
Response to Reviewer wjK1 (1/6)
W1: Correction to Table 8 and channel projection in zero-shot evaluation
Correction to Table 8 and 10
Thank you for flagging this discrepancy. Tables 8 and 10 evaluate Ego + AUX1. We mistakenly entered ALF scores indexed by AUX2. The correct scores correspond to the AUX1 configuration used at inference. When the receiving model and AUX1 are fixed, the ALF score is identical across rows that list different AUX2 configurations. In the revised manuscript, we will correct the ALF entries in Tables 8 and 10 to report the scores for the evaluated AUX1 configurations.
Clarification on channel projection
In zero-shot evaluation, an unseen AUX encoder can produce a bird’s-eye-view (BEV) feature map with a different channel count from what the pretrained receiving module expects. A channel mismatch prevents the module from processing the input directly. To run the pretrained baseline without retraining, we apply a fixed 1×1 convolution that maps the AUX feature vector at each BEV grid cell to the expected channel count. We bypass this projection when the channel counts match.
We considered channel truncation, zero-padding, and an unconstrained randomly initialized 1×1 convolution. Truncation discards channels by index, zero-padding adds empty channels, and an unconstrained random projection can scale feature directions unevenly. We chose orthogonal initialization as a fairer training-free control because it mixes input channels without selecting them by index and scales the retained directions uniformly. The projection weights remain fixed during zero-shot evaluation. The effect of this projection on AP in channel-mismatched cases remains unquantified.
For transfers between PP6 and PP8, all four feature-fusion baselines bypass the added projection in both directions. For transfers between SD2 and SD3, E2E and CodeFilling project the AUX features from 512 to 256 channels to match the PP4 Ego features. MPDA’s resizer and GenComm’s message encoder expect the channel count used during their training. In both transfer directions, the incoming features and the pretrained AUX modules use 512 channels, so MPDA and GenComm bypass the added projection. We will clarify these conditions in Appendix B.5.