Parameter-Inverted Image Pyramid Networks

Image pyramids are commonly used in modern computer vision tasks to obtain multi-scale features for precise understanding of images. However, image pyramids process multiple resolutions of images using the same large-scale model, which requires significant computational cost. To overcome this issue, we propose a novel network architecture known as the Parameter-Inverted Image Pyramid Networks (PIIP). Our core idea is to use models with different parameter sizes to process different resolution levels of the image pyramid, thereby balancing computational efficiency and performance. Specifically, the input to PIIP is a set of multi-scale images, where higher resolution images are processed by smaller networks. We further propose a feature interaction mechanism to allow features of different resolutions to complement each other and effectively integrate information from different spatial scales. Extensive experiments demonstrate that the PIIP achieves superior performance in tasks such as object detection, segmentation, and image classification, compared to traditional image pyramid methods and single-branch networks, while reducing computational cost. Notably, when applying our method on a large-scale vision foundation model InternViT-6B, we improve its performance by 1%-2% on detection and segmentation with only 40%-60% of the original computation. These results validate the effectiveness of the PIIP approach and provide a new technical direction for future vision computing tasks. Our code and models are available at https://github.com/OpenGVLab/PIIP.

Paper

References (68)

Scroll for more · 38 remaining

Similar papers

Peer review

Reviewer hVKM9/10 · confidence 5/52024-07-07

Summary

This paper presents a novel method named parameter-inverted image pyramid to handle the issues of high computation overhead of image pyramids. It uses different model sizes to process different resolutions. The method achieves significant results on tasks of object detection, segmentation, and classification by reducing overhead and improving performance.

Strengths

1. The paper is well-organized and effectively presents its content, making for a clear and coherent read. 2. The idea of Parameter-Inverted, e.g., larger models for small images and smaller models for large images, sounds interesting. 3. The method is novel and simple without complicated handcraft structures. The direct use of existing ViT models is especially interesting and can be easily extended to other tasks. 4. The experiments are abundant and convincing. The authors provide sufficient experimental results with tables and scatter plots to verify that the proposed framework is solid and effective. The performance on 6B models is impressive.

Weaknesses

1. The authors did not specify whether the FLOPs and Param in Table 1,3 include a branch merging module. 2. The discussion with other multi-resolution networks is not sufficient, e.g. HRNet [16,45,57] 3. The proposed framework can be verified on stronger pre-training, e.g. ViTDet-L (BEiTv2) vs. PIIP-SBL (BEiTv2), to further enhance the effectiveness of the method. 4. In Lines 180-181, the authors used layer-wise learning rate decay, but the authors do not indicate how to deal with ViT combinations with different numbers of layers, such as PIIP-SBL, ViT-S/B has 12 layers, and ViT-L has 24 layers. 5. The text in Figure 1 could be slightly larger.

Questions

1. I am just a little curious, for PIIP-TSB, why ViT-T from DeiT but ViT-S/B from DeiT-III? Won't different pre-training methods have some impact? According to Table 5, ViT-T with large resolution input plays the most important role, so the weaker pre-training ViT-T should limit the performance of the framework? 2. (Unimportant) In Figure 4, why are the improvements in detection better than those on instance segmentation?

Rating

9

Confidence

5

Soundness

4

Presentation

4

Contribution

4

Limitations

NA

Reviewer Lww47/10 · confidence 4/52024-07-13

Summary

This paper proposes two techniques: 1) Models with different parameter sizes to process different resolution levels of the image pyramid. 2) A feature interaction mechanism to integrate information from different spatial scales. Extensive experiment results are used to support the claims.

Strengths

Both techniques are sound. Experiments are well documented and extensive.

Weaknesses

It is unclear to me if the better accuracy in table 1, 2, 3 and 4 are entirely due to the proposed techniques or it could also be partially explained by the fact that the highest resolution in the pyramid is larger. Take table 1 as an example, the input resolution for baseline is 1024. But the highest resolution for PIIP is 1792. What if you use 1792 for baseline? It is perfectly fine to first hold FLOPS constant and check accuracy, as shown in table-1. But you should also hold input constant and check accuracy.

Questions

See above

Rating

7

Confidence

4

Soundness

3

Presentation

4

Contribution

3

Limitations

yes

Reviewer H3y67/10 · confidence 4/52024-07-15

Summary

The authors propose a novel vision architecture, Parameter-Inverted Image Pyramid Networks (PIIP), which can be applied to different tasks, including classification, object detection, and instance or semantic segmentation. The authors aim to take advantage of the multi-scale information of image pyramids, without the usual excessive computational demands of processing an image in multiple different resolutions. The core idea is to use a family of networks of different computational requirements, e.g., ViT-S/B/L, and apply the lightest model (least parameters) to the image of the highest resolution in the pyramid, apply the second lighter model to the next bigger image, and so on, ending up to apply the heaviest model (most parameters) to the image of the smallest resolution. The intuition is that stronger models with more parameters can be used to extract semantically rich contextual features from images of coarser resolution, while lighter models can be efficiently applied to extract low-level features from high-resolution images. This way, the computational requirements of the different models are balanced, meaningful features from all scales are extracted, and images can be processed in higher resolution because the lighter and less expensive models are processing them. The models are applied to the different images in the pyramid in parallel, and as features are extracted, the models communicate with each other through a proposed Interaction Unit. This way, features of different semantic meaning extracted from different models complement and inform each other. After the features from each individual model are extracted, they are merged to be used for the task at hand. The authors perform extensive evaluation of PIIP on 4 different tasks, object detection, instance segmentation, semantic segmentation, and image classification. The authors initialize the PIIP backbones with pre-trained ViT models, and for each task they fine-tune appropriate heads. PIIP is compared to baselines of different scale, and in all experiments it demonstrates superior or comparable performance. In addition, the authors perform multiple ablations, offering insights about the behavior of PIIP, and the importance of different design choices.

Strengths

Originality: 1. The core idea of using a parameter-inverted pyramid seemed to me counter-intuitive at first, since I would expect heavier models to be applied to the images of higher resolution, which contain more information. However, the authors make a good argument about why the parameter-inverted pyramid design is sensible, and I think it is a novel idea, that in its simplicity offers a valuable contribution to the community. Quality: 1. The authors provide extensive experiments on multiple tasks, and in their experiments they control for the computational requirements of the models, offering meaningful comparisons. 2. The experimental evaluation includes multiple ablations, which shed light on different aspects of the proposed architecture. Clarity: 1. The manuscript is well written, and easy to follow. The authors explain the intuition and the specifics of their method in detail, accompanying the text with clear visualizations. Significance: 1. In addition to the novelty of the proposed idea of the parameter-inverted pyramid, I think a useful conclusion of the paper is the importance of using higher resolution inputs, even if they are processed with models of smaller size. To my understanding, this is the main reason PIIP outperforms the baselines, and is highlighted by the authors in Section 4.5 (ln 260 - 266). The importance of the input resolution is not something new, but the authors offer more evidence about its impact, and importantly, they offer a new paradigm to take advantage of it.

Weaknesses

Quality: 1. My main concern about this work is that the authors don’t provide actual timings and memory measurements. Computational requirements are quantified through FLOPs, however, theoretical gains in FLOPs don’t always translate to benefits in practice when the compared architectures have considerably different designs. For example, the PIIP models may be applied in parallel, but the interactions between the features may force the models to wait for each other, adding a sequential element that may add to the latency. I want to clarify that I am not claiming that this is the case, but that actual timings beyond theoretical FLOPs are needed to offer conclusive evidence, especially when efficiency is one of the main claims of this work. 2. About memory requirements, it seems to me that using a whole family of models may be prohibitive in many settings, so, memory should be explicitly measured. In addition, the authors control in their experiments for FLOPs, e.g., in Table 1 they show how PIIP models perform compared to baselines with similar FLOPs. I think it would be useful to control for memory too, e.g., in Table 1 the best PIIP models have more than 50% higher number of parameters compared to the baselines. Based on this, I think a question that naturally arises is what if the baseline was a larger model with equal parameters? If we have an extra memory budget, is PIIP the best way to use it? I would like to clarify that I don’t think the authors should necessarily have additional experiments controlling for memory, but if it is a disadvantage of PIIP, as it seems to be due to the parameter count, I think they should discuss it in the limitations. Clarity: 1. In Table 2, the results for ViTDet-B and ViTDet-L are different compared to Table 1, why is that? Similarly, in Table 2, the results of PIIP-TSB and PIIP-SBL are different compared to Table 1. For the PIIP-SBL model, I see that the scores in Table 2 match the scores of the best model reported in Table 12, where models use higher resolutions, is this the reason for the discrepancy? However, the reported performance of the PIIP-TSB in Table 2 is higher than any performance reported for PIIP-TSB in Table 12, so, how the performance reported in Table 2 is achieved? I think the authors should clarify the configurations of the models they use in their experiments. 2. Some of the Table captions are not informative enough. For example, it is not clear why crop size is underlined in Tables 3, 6, 7. On a similar note, in the ablation about branch merging (ln 267 - 269, and Table 5), the authors don’t mention the dataset they use. I guess it is MS COCO, but I think it should be explicitly mentioned. 3. In the caption of Table 8, it is mentioned “‘PI’ and ‘IP’, ‘Inter.’ represent parameter-inverted, image pyramid and interactions”, I think it should be “‘PI’, ‘IP’, and ‘Inter.’”. 4. In Ln 262 the authors mention “green dashed circle in Fig. 5(a)”, but I think it should be “blue dashed circle”. Significance: 1. PIIP requires a family of backbones for its processing, and the authors use pre-trained models in their experiments, while they also include some experiments on training from scratch in the Appendix. I think the need to train multiple backbones, or the need to find multiple pre-train models, may be an undesirable overhead for the adoption of the method.

Questions

1. In Section 4.1 the authors provide the pre-trained architectures they used for different tasks, how did they decide which models to use? 2. What is the formula the authors use to calculate the FLOPs of the models? 3. Why there are no comparisons with state-of-the-art models on image classification? 4. In Table 8, what is the range of resolutions that the model in the second row is trained on?

Rating

7

Confidence

4

Soundness

3

Presentation

3

Contribution

3

Limitations

The authors offer a short discussion of the limitations in Section 5, which I think should be expanded to address the memory requirements of PIIP.

Reviewer H3y62024-08-11

Thank you for your reply

I would like to thank the authors for their detailed response. As I mentioned in my review, my main concerns were about actual timings and memory. About timings, the numbers the authors provide are encouraging, and I agree that total agreement between FLOPs and actual runtime is not necessary from the very beginning, when a method is not yet fully optimized. I would suggest the authors include this information in Future Work or Limitations. About memory, I agree that memory is not easy to control with pre-trained models, and some of the existing results allude to favorable comparisons of PIIP with baselines of similar memory. So, I find the authors' response sensible and encouraging, but in the absence of more detailed resource comparisons in the current manuscript, I will maintain my initial score.

Authorsrebuttal2024-08-13

Thank you for your valuable discussion and positive decision. We will make corresponding revisions in the final manuscript.

Reviewer Lww42024-08-13

Thanks for the new results. I will keep my positive rating.

Authorsrebuttal2024-08-13

Thank you for your valuable discussion and positive decision. We will make corresponding revisions in the final manuscript.

Reviewer hVKM2024-08-13

I agree with reviewer H3y6 that the parameter-inverted design is initially counter-intuitive but impressive in the end, given the experimental results. Since my concerns and questions have been addressed by the authors, and considering the high quality of this research has been recognized by all the reviewers, I keep my positive rating.

Program Chairsdecision2024-09-25

Decision

Accept (spotlight)

© 2026 NYSGPT2525 LLC