Radomir Mech
Research Subjects
Radomir Mech has a total of 69 co-authors. Here are the top co-authors by number of shared publications:
- Zhe Lin
Adobe Research, San Jose, USA
5 shared publications - Bedrich Benes
Purdue University, West Lafayette, USA
4 shared publications - Xiaohui Shen
Adobe Research
4 shared publications - Nathan Carr
Adobe Research, San Jose, USA
3 shared publications - Gavin Stuart Peter Miller
Adobe Research, San Jose, United States of America
2 shared publications - Jianming Zhang
Adobe Research
2 shared publications
Latest Publications
- Open Access
3D Space as a Scratchpad for Editable Text-to-Image Generation
Recent progress in large language models (LLMs) has shown that reasoning improves when intermediate thoughts are externalized into explicit workspaces, such as chain-of-thought traces or tool-augmented reasoning. Yet, visual language models (VLMs) lack an analogous mechanism for spatial reasoning, limiting their ability to generate images that accurately reflect geometric relations, object identities, and compositional intent. We introduce the concept of a spatial scratchpad -- a 3D reasoning substrate that bridges linguistic intent and image synthesis. Given a text prompt, our framework parses subjects and background elements, instantiates them as editable 3D meshes, and employs agentic scene planning for placement, orientation, and viewpoint selection. The resulting 3D arrangement is rendered back into the image domain with identity-preserving cues, enabling the VLM to generate spatially consistent and visually coherent outputs. Unlike prior 2D layout-based methods, our approach supports intuitive 3D edits that propagate reliably into final images. Empirically, it achieves a 32% improvement in text alignment on GenAI-Bench, demonstrating the benefit of explicit 3D reasoning for precise, controllable image generation. Our results highlight a new paradigm for vision-language models that deliberate not only in language, but also in space. Code and visualizations at https://oindrilasaha.github.io/3DScratchpad/
Open Access3D Space as a Scratchpad for Editable Text-to-Image Generation
Recent progress in large language models (LLMs) has shown that reasoning improves when intermediate thoughts are externalized into explicit workspaces, such as chain-of-thought traces or tool-augmented reasoning. Yet, visual language models (VLMs) lack an analogous mechanism for spatial reasoning, limiting their ability to generate images that accurately reflect geometric relations, object identities, and compositional intent. We introduce the concept of a spatial scratchpad -- a 3D reasoning substrate that bridges linguistic intent and image synthesis. Given a text prompt, our framework parses subjects and background elements, instantiates them as editable 3D meshes, and employs agentic scene planning for placement, orientation, and viewpoint selection. The resulting 3D arrangement is rendered back into the image domain with identity-preserving cues, enabling the VLM to generate spatially consistent and visually coherent outputs. Unlike prior 2D layout-based methods, our approach supports intuitive 3D edits that propagate reliably into final images. Empirically, it achieves a 32% improvement in text alignment on GenAI-Bench, demonstrating the benefit of explicit 3D reasoning for precise, controllable image generation. Our results highlight a new paradigm for vision-language models that deliberate not only in language, but also in space. Code and visualizations at https://oindrilasaha.github.io/3DScratchpad/
- Open Access
SIGMA-GEN: Structure and Identity Guided Multi-subject Assembly for Image Generation
We present SIGMA-GEN, a unified framework for multi-identity preserving image generation. Unlike prior approaches, SIGMA-GEN is the first to enable single-pass multi-subject identity-preserved generation guided by both structural and spatial constraints. A key strength of our method is its ability to support user guidance at various levels of precision -- from coarse 2D or 3D boxes to pixel-level segmentations and depth -- with a single model. To enable this, we introduce SIGMA-SET27K, a novel synthetic dataset that provides identity, structure, and spatial information for over 100k unique subjects across 27k images. Through extensive evaluation we demonstrate that SIGMA-GEN achieves state-of-the-art performance in identity preservation, image generation quality, and speed. Code and visualizations at https://oindrilasaha.github.io/SIGMA-Gen/
Open AccessSIGMA-GEN: Structure and Identity Guided Multi-subject Assembly for Image Generation
We present SIGMA-GEN, a unified framework for multi-identity preserving image generation. Unlike prior approaches, SIGMA-GEN is the first to enable single-pass multi-subject identity-preserved generation guided by both structural and spatial constraints. A key strength of our method is its ability to support user guidance at various levels of precision -- from coarse 2D or 3D boxes to pixel-level segmentations and depth -- with a single model. To enable this, we introduce SIGMA-SET27K, a novel synthetic dataset that provides identity, structure, and spatial information for over 100k unique subjects across 27k images. Through extensive evaluation we demonstrate that SIGMA-GEN achieves state-of-the-art performance in identity preservation, image generation quality, and speed. Code and visualizations at https://oindrilasaha.github.io/SIGMA-Gen/
- Open Access
DISN: Deep Implicit Surface Network for High-quality Single-view 3D Reconstruction
Reconstructing 3D shapes from single-view images has been a long-standing research problem. In this paper, we present DISN, a Deep Implicit Surface Network which can generate a high-quality detail-rich 3D mesh from an 2D image by predicting the underlying signed distance fields. In addition to utilizing global image features, DISN predicts the projected location for each 3D point on the 2D image, and extracts local features from the image feature maps. Combining global and local features significantly improves the accuracy of the signed distance field prediction, especially for the detail-rich areas. To the best of our knowledge, DISN is the first method that constantly captures details such as holes and thin structures present in 3D shapes from single-view images. DISN achieves the state-of-the-art single-view reconstruction performance on a variety of shape categories reconstructed from both synthetic and real images. Code is available at https://github.com/xharlie/DISN The supplementary can be found at https://xharlie.github.io/images/neurips_2019_supp.pdf
Open AccessDISN: Deep Implicit Surface Network for High-quality Single-view 3D Reconstruction
Reconstructing 3D shapes from single-view images has been a long-standing research problem. In this paper, we present DISN, a Deep Implicit Surface Network which can generate a high-quality detail-rich 3D mesh from an 2D image by predicting the underlying signed distance fields. In addition to utilizing global image features, DISN predicts the projected location for each 3D point on the 2D image, and extracts local features from the image feature maps. Combining global and local features significantly improves the accuracy of the signed distance field prediction, especially for the detail-rich areas. To the best of our knowledge, DISN is the first method that constantly captures details such as holes and thin structures present in 3D shapes from single-view images. DISN achieves the state-of-the-art single-view reconstruction performance on a variety of shape categories reconstructed from both synthetic and real images. Code is available at https://github.com/xharlie/DISN The supplementary can be found at https://xharlie.github.io/images/neurips_2019_supp.pdf
- 6 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 30 (8) , 4514-4526
ANISE: Assembly-Based Neural Implicit Surface Reconstruction
We present ANISE, a method that reconstructs a 3D shape from partial observations (images or sparse point clouds) using a part-aware neural implicit shape representation. The shape is formulated as an assembly of neural implicit functions, each representing a different part instance. In contrast to previous approaches, the prediction of this representation proceeds in a coarse-to-fine manner. Our model first reconstructs a structural arrangement of the shape in the form of geometric transformations of its part instances. Conditioned on them, the model predicts part latent codes encoding their surface geometry. Reconstructions can be obtained in two ways: (i) by directly decoding the part latent codes to part implicit functions, then combining them into the final shape; or (ii) by using part latents to retrieve similar part instances in a part database and assembling them in a single shape. We demonstrate that, when performing reconstruction by decoding part representations into implicit functions, our method achieves state-of-the-art part-aware reconstruction results from both images and sparse point clouds. When reconstructing shapes by assembling parts retrieved from a dataset, our approach significantly outperforms traditional shape retrieval methods even when significantly restricting the database size. We present our results in well-known sparse point cloud reconstruction and single-view reconstruction benchmarks.
- 6 April 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 30 (8) , 4514-4526
ANISE: Assembly-Based Neural Implicit Surface Reconstruction
We present ANISE, a method that reconstructs a 3D shape from partial observations (images or sparse point clouds) using a part-aware neural implicit shape representation. The shape is formulated as an assembly of neural implicit functions, each representing a different part instance. In contrast to previous approaches, the prediction of this representation proceeds in a coarse-to-fine manner. Our model first reconstructs a structural arrangement of the shape in the form of geometric transformations of its part instances. Conditioned on them, the model predicts part latent codes encoding their surface geometry. Reconstructions can be obtained in two ways: (i) by directly decoding the part latent codes to part implicit functions, then combining them into the final shape; or (ii) by using part latents to retrieve similar part instances in a part database and assembling them in a single shape. We demonstrate that, when performing reconstruction by decoding part representations into implicit functions, our method achieves state-of-the-art part-aware reconstruction results from both images and sparse point clouds. When reconstructing shapes by assembling parts retrieved from a dataset, our approach significantly outperforms traditional shape retrieval methods even when significantly restricting the database size. We present our results in well-known sparse point cloud reconstruction and single-view reconstruction benchmarks.
- Open Access
MatFormer: A Generative Model for Procedural Materials
Procedural material graphs are a compact, parameteric, and resolution-independent representation that are a popular choice for material authoring. However, designing procedural materials requires significant expertise and publicly accessible libraries contain only a few thousand such graphs. We present MatFormer, a generative model that can produce a diverse set of high-quality procedural materials with complex spatial patterns and appearance. While procedural materials can be modeled as directed (operation) graphs, they contain arbitrary numbers of heterogeneous nodes with unstructured, often long-range node connections, and functional constraints on node parameters and connections. MatFormer addresses these challenges with a multi-stage transformer-based model that sequentially generates nodes, node parameters, and edges, while ensuring the semantic validity of the graph. In addition to generation, MatFormer can be used for the auto-completion and exploration of partial material graphs. We qualitatively and quantitatively demonstrate that our method outperforms alternative approaches, in both generated graph and material quality.
Open AccessMatFormer: A Generative Model for Procedural Materials
Procedural material graphs are a compact, parameteric, and resolution-independent representation that are a popular choice for material authoring. However, designing procedural materials requires significant expertise and publicly accessible libraries contain only a few thousand such graphs. We present MatFormer, a generative model that can produce a diverse set of high-quality procedural materials with complex spatial patterns and appearance. While procedural materials can be modeled as directed (operation) graphs, they contain arbitrary numbers of heterogeneous nodes with unstructured, often long-range node connections, and functional constraints on node parameters and connections. MatFormer addresses these challenges with a multi-stage transformer-based model that sequentially generates nodes, node parameters, and edges, while ensuring the semantic validity of the graph. In addition to generation, MatFormer can be used for the auto-completion and exploration of partial material graphs. We qualitatively and quantitatively demonstrate that our method outperforms alternative approaches, in both generated graph and material quality.
- Open Access
- 22 July 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 41 (4) , 1-12
MatFormer
Procedural material graphs are a compact, parameteric, and resolution-independent representation that are a popular choice for material authoring. However, designing procedural materials requires significant expertise and publicly accessible libraries contain only a few thousand such graphs. We present MatFormer, a generative model that can produce a diverse set of high-quality procedural materials with complex spatial patterns and appearance. While procedural materials can be modeled as directed (operation) graphs, they contain arbitrary numbers of heterogeneous nodes with unstructured, often long-range node connections, and functional constraints on node parameters and connections. MatFormer addresses these challenges with a multi-stage transformer-based model that sequentially generates nodes, node parameters, and edges, while ensuring the semantic validity of the graph. In addition to generation, MatFormer can be used for the auto-completion and exploration of partial material graphs. We qualitatively and quantitatively demonstrate that our method outperforms alternative approaches, in both generated graph and material quality.
- 22 July 2022
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Graphics
- Vol. 41 (4) , 1-12
Open AccessMatFormer
Procedural material graphs are a compact, parameteric, and resolution-independent representation that are a popular choice for material authoring. However, designing procedural materials requires significant expertise and publicly accessible libraries contain only a few thousand such graphs. We present MatFormer, a generative model that can produce a diverse set of high-quality procedural materials with complex spatial patterns and appearance. While procedural materials can be modeled as directed (operation) graphs, they contain arbitrary numbers of heterogeneous nodes with unstructured, often long-range node connections, and functional constraints on node parameters and connections. MatFormer addresses these challenges with a multi-stage transformer-based model that sequentially generates nodes, node parameters, and edges, while ensuring the semantic validity of the graph. In addition to generation, MatFormer can be used for the auto-completion and exploration of partial material graphs. We qualitatively and quantitatively demonstrate that our method outperforms alternative approaches, in both generated graph and material quality.
- 24 May 2022
- Published by Wiley in Computer Graphics Forum
- Vol. 41 (2) , 289-307
Automatic Differentiable Procedural Modeling
- 24 May 2022
- Published by Wiley in Computer Graphics Forum
- Vol. 41 (2) , 289-307
Automatic Differentiable Procedural Modeling
- 1 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 28 (4) , 1758-1772
FAME: 3D Shape Generation via Functionality-Aware Model Evolution
We introduce a modeling tool which can evolve a set of 3D objects in a functionality-aware manner. Our goal is for the evolution to generate large and diverse sets of plausible 3D objects for data augmentation, constrained modeling, as well as open-ended exploration to possibly inspire new designs. Starting with an initial population of 3D objects belonging to one or more functional categories, we evolve the shapes through part recombination to produce generations of hybrids or crossbreeds between parents from the heterogeneous shape collection. Evolutionary selection of offsprings is guided both by a functional plausibility score derived from functionality analysis of shapes in the initial population and user preference, as in a design gallery. Since cross-category hybridization may result in offsprings not belonging to any of the known functional categories, we develop a means for functionality partial matching to evaluate functional plausibility on partial shapes. We show a variety of plausible hybrid shapes generated by our functionality-aware model evolution, which can complement existing datasets as training data and boost the performance of contemporary data-driven segmentation schemes, especially in challenging cases. Our tool supports constrained modeling, allowing users to restrict or steer the model evolution with functionality labels. At the same time, unexpected yet functional object prototypes can emerge during open-ended exploration owing to structure breaking when evolving a heterogeneous collection.
- 1 April 2022
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 28 (4) , 1758-1772
FAME: 3D Shape Generation via Functionality-Aware Model Evolution
We introduce a modeling tool which can evolve a set of 3D objects in a functionality-aware manner. Our goal is for the evolution to generate large and diverse sets of plausible 3D objects for data augmentation, constrained modeling, as well as open-ended exploration to possibly inspire new designs. Starting with an initial population of 3D objects belonging to one or more functional categories, we evolve the shapes through part recombination to produce generations of hybrids or crossbreeds between parents from the heterogeneous shape collection. Evolutionary selection of offsprings is guided both by a functional plausibility score derived from functionality analysis of shapes in the initial population and user preference, as in a design gallery. Since cross-category hybridization may result in offsprings not belonging to any of the known functional categories, we develop a means for functionality partial matching to evaluate functional plausibility on partial shapes. We show a variety of plausible hybrid shapes generated by our functionality-aware model evolution, which can complement existing datasets as training data and boost the performance of contemporary data-driven segmentation schemes, especially in challenging cases. Our tool supports constrained modeling, allowing users to restrict or steer the model evolution with functionality labels. At the same time, unexpected yet functional object prototypes can emerge during open-ended exploration owing to structure breaking when evolving a heterogeneous collection.
- 18 May 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 27 (10) , 3968-3981
PICO: Procedural Iterative Constrained Optimizer for Geometric Modeling
Procedural modeling has produced amazing results, yet fundamental issues such as controllability and limited user guidance persist. We introduce a novel procedural model called PICO (Procedural Iterative Constrained Optimizer) and PICO-Graph that is the underlying procedural model designed with optimization in mind. The key novelty of PICO is that it enables the exploration of generative designs by combining both user and environmental constraints into a single framework by using optimization without the need to write procedural rules. The PICO-Graph procedural model consists of a set of geometry generating operations and a set of axioms connected in a directed cyclic graph. The forward generation is initiated by a set of axioms that use the connections to send coordinate systems and geometric objects through the PICO-Graph, which in turn generates more objects. This allows for fast generation of complex and varied geometries. Moreover, we combine PICO-Graph with efficient optimization that allows for quick exploration of the generated models and the generation of variants. The user defines the rules, the axioms, and the set of constraints; for example, whether an existing object should be supported by the generated model, whether symmetries exist, whether the object should spin, etc. PICO then generates a class of geometric models and optimizes them so that they fulfill the constraints. The generation and the optimization in our implementation provides interactive user control during model execution providing continuous feedback. For example, the user can sketch the constraints and guide the geometry to meet these specified goals. We show PICO on a variety of examples such as the generation of procedural chairs with multiple supports, generation of support structures for 3D printing, generation of spinning objects, or generation of procedural terrains matching a given input. Our framework could be used as a component in a larger design workflow; its strongest application is in the early rapid ideation and prototyping phases.
- 18 May 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 27 (10) , 3968-3981
PICO: Procedural Iterative Constrained Optimizer for Geometric Modeling
Procedural modeling has produced amazing results, yet fundamental issues such as controllability and limited user guidance persist. We introduce a novel procedural model called PICO (Procedural Iterative Constrained Optimizer) and PICO-Graph that is the underlying procedural model designed with optimization in mind. The key novelty of PICO is that it enables the exploration of generative designs by combining both user and environmental constraints into a single framework by using optimization without the need to write procedural rules. The PICO-Graph procedural model consists of a set of geometry generating operations and a set of axioms connected in a directed cyclic graph. The forward generation is initiated by a set of axioms that use the connections to send coordinate systems and geometric objects through the PICO-Graph, which in turn generates more objects. This allows for fast generation of complex and varied geometries. Moreover, we combine PICO-Graph with efficient optimization that allows for quick exploration of the generated models and the generation of variants. The user defines the rules, the axioms, and the set of constraints; for example, whether an existing object should be supported by the generated model, whether symmetries exist, whether the object should spin, etc. PICO then generates a class of geometric models and optimizes them so that they fulfill the constraints. The generation and the optimization in our implementation provides interactive user control during model execution providing continuous feedback. For example, the user can sketch the constraints and guide the geometry to meet these specified goals. We show PICO on a variety of examples such as the generation of procedural chairs with multiple supports, generation of support structures for 3D printing, generation of spinning objects, or generation of procedural terrains matching a given input. Our framework could be used as a component in a larger design workflow; its strongest application is in the early rapid ideation and prototyping phases.
- 10 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 43 (3) , 1009-1021
Sequence-to-Segments Networks for Detecting Segments in Videos
Detecting segments of interest from videos is a common problem for many applications. And yet it is a challenging problem as it often requires not only knowledge of individual target segments, but also contextual understanding of the entire video and the relationships between the target segments. To address this problem, we propose the Sequence-to-Segments Network (S 2 N), a novel and general end-to-end sequential encoder-decoder architecture. S 2 N first encodes the input video into a sequence of hidden states that capture information progressively, as it appears in the video. It then employs the Segment Detection Unit (SDU), a novel decoding architecture, that sequentially detects segments. At each decoding step, the SDU integrates the decoder state and encoder hidden states to detect a target segment. During training, we address the problem of finding the best assignment of predicted segments to ground truth using the Hungarian Matching Algorithm with Lexicographic Cost. Additionally we propose to use the squared Earth Mover's Distance to optimize the localization errors of the segments. We show the state-of-the-art performance of S 2 N across numerous tasks, including video highlighting, video summarization, and human action proposal generation.
- 10 September 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Pattern Analysis and Machine Intelligence
- Vol. 43 (3) , 1009-1021
Sequence-to-Segments Networks for Detecting Segments in Videos
Detecting segments of interest from videos is a common problem for many applications. And yet it is a challenging problem as it often requires not only knowledge of individual target segments, but also contextual understanding of the entire video and the relationships between the target segments. To address this problem, we propose the Sequence-to-Segments Network (S 2 N), a novel and general end-to-end sequential encoder-decoder architecture. S 2 N first encodes the input video into a sequence of hidden states that capture information progressively, as it appears in the video. It then employs the Segment Detection Unit (SDU), a novel decoding architecture, that sequentially detects segments. At each decoding step, the SDU integrates the decoder state and encoder hidden states to detect a target segment. During training, we address the problem of finding the best assignment of predicted segments to ground truth using the Hungarian Matching Algorithm with Lexicographic Cost. Additionally we propose to use the squared Earth Mover's Distance to optimize the localization errors of the segments. We show the state-of-the-art performance of S 2 N across numerous tasks, including video highlighting, video summarization, and human action proposal generation.