Niladri Shekhar Dutt
@niladridutt.bsky.social
Research Intern @adobe.com | PhD @ucl.ac.uk | @ellis.eu | ex-Nvidia, Berkeley | Interested in generative modelling in vision and graphics + reasoning (LLMs)
https://niladridutt.com/
https://niladridutt.com/
🧵10/10 Lastly, huge thanks to my co-advisors Niloy and Duygu!
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
MonetGPT
monetgpt.github.io
May 27, 2025 at 3:13 PM
🧵10/10 Lastly, huge thanks to my co-advisors Niloy and Duygu!
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
🧵9/10 We quantitaively evaluate on the Adobe5k dataset as well as conduct user studies by expert and novice users. Our evaluations show that MonetGPT outperforms open-source alternatives and performs comparably to Google Photos AutoEnhance (closed-source).
May 27, 2025 at 3:13 PM
🧵9/10 We quantitaively evaluate on the Adobe5k dataset as well as conduct user studies by expert and novice users. Our evaluations show that MonetGPT outperforms open-source alternatives and performs comparably to Google Photos AutoEnhance (closed-source).
🧵8/10 Photo editing is subjective 🎨. Our framework adapts to user preference by guidance from natural language tags like ‘vibrant’ or ‘retro vibe’ to produce personalized and stylistically distinct retouching plans from the same input image.
May 27, 2025 at 3:13 PM
🧵8/10 Photo editing is subjective 🎨. Our framework adapts to user preference by guidance from natural language tags like ‘vibrant’ or ‘retro vibe’ to produce personalized and stylistically distinct retouching plans from the same input image.
🧵7/10 Our puzzle-based training with a 'reasoning as a pathway' approach allows MonetGPT to generate detailed justifications for each edit, delivering truly explainable image retouching
May 27, 2025 at 3:13 PM
🧵7/10 Our puzzle-based training with a 'reasoning as a pathway' approach allows MonetGPT to generate detailed justifications for each edit, delivering truly explainable image retouching
🧵6/10 🧩 Puzzle C builds planning capabilities. The model learns to generate a complete, multi-step retouching plan to enhance a photo, structuring its reasoning as a sequence of discrete issues and solutions for clarity and control.
May 27, 2025 at 3:13 PM
🧵6/10 🧩 Puzzle C builds planning capabilities. The model learns to generate a complete, multi-step retouching plan to enhance a photo, structuring its reasoning as a sequence of discrete issues and solutions for clarity and control.
🧵5/10 🧩 Puzzle B imparts aesthetic judgement. By ranking professionally edited photos against altered versions, the MLLM learns to recognize the visual characteristics of an optimally adjusted image for any given operation, building an internal aesthetic model.
May 27, 2025 at 3:13 PM
🧵5/10 🧩 Puzzle B imparts aesthetic judgement. By ranking professionally edited photos against altered versions, the MLLM learns to recognize the visual characteristics of an optimally adjusted image for any given operation, building an internal aesthetic model.
🧵4/10 🧩 Puzzle A builds an understanding of individual operations. The MLLM learns to map visual changes in before/after images to a specific tool and its precise parameter value, effectively learning the semantics of our procedural library.
May 27, 2025 at 3:13 PM
🧵4/10 🧩 Puzzle A builds an understanding of individual operations. The MLLM learns to map visual changes in before/after images to a specific tool and its precise parameter value, effectively learning the semantics of our procedural library.
🧵3/10 Our key recipe: MLLMs struggle to predict edit values directly. We solve this by generating rich textual reasoning for each puzzle ✍️. We then fine-tune MonetGPT on this data, creating a 'reasoning pathway' that enables it to regress final adjustment values.
May 27, 2025 at 3:13 PM
🧵3/10 Our key recipe: MLLMs struggle to predict edit values directly. We solve this by generating rich textual reasoning for each puzzle ✍️. We then fine-tune MonetGPT on this data, creating a 'reasoning pathway' that enables it to regress final adjustment values.
🧵2/10 MLLMs lack the visual understanding to plan edits. 🧠 So, we use expert photos as our ground truth and work backward, procedurally creating puzzles by assuming any change to an expert edit makes it less optimal
May 27, 2025 at 3:13 PM
🧵2/10 MLLMs lack the visual understanding to plan edits. 🧠 So, we use expert photos as our ground truth and work backward, procedurally creating puzzles by assuming any change to an expert edit makes it less optimal
🧵10/10 Lastly, huge thanks to my co-advisors Niloy and Duygu!
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
MonetGPT
monetgpt.github.io
May 27, 2025 at 3:04 PM
🧵10/10 Lastly, huge thanks to my co-advisors Niloy and Duygu!
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
For more details check out our paper below-
🌐 Project Website: monetgpt.github.io
📄 Arxiv: arxiv.org/abs/2505.06176
🧵9/10 We quantitaively evaluate on the Adobe5k dataset as well as conduct user studies by expert and novice users. Our evaluations show that MonetGPT outperforms open-source alternatives and performs comparably to Google Photos AutoEnhance (closed-source).
May 27, 2025 at 3:04 PM
🧵9/10 We quantitaively evaluate on the Adobe5k dataset as well as conduct user studies by expert and novice users. Our evaluations show that MonetGPT outperforms open-source alternatives and performs comparably to Google Photos AutoEnhance (closed-source).
🧵8/10 Photo editing is subjective 🎨. Our framework adapts to user preference by guidance from natural language tags like ‘vibrant’ or ‘retro vibe’ to produce personalized and stylistically distinct retouching plans from the same input image.
May 27, 2025 at 3:04 PM
🧵8/10 Photo editing is subjective 🎨. Our framework adapts to user preference by guidance from natural language tags like ‘vibrant’ or ‘retro vibe’ to produce personalized and stylistically distinct retouching plans from the same input image.
🧵7/10 Our puzzle-based training with a 'reasoning as a pathway' approach allows MonetGPT to generate detailed justifications for each edit, delivering truly explainable image retouching
May 27, 2025 at 3:04 PM
🧵7/10 Our puzzle-based training with a 'reasoning as a pathway' approach allows MonetGPT to generate detailed justifications for each edit, delivering truly explainable image retouching
🧵6/10 🧩 Puzzle C builds planning capabilities. The model learns to generate a complete, multi-step retouching plan to enhance a photo, structuring its reasoning as a sequence of discrete issues and solutions for clarity and control.
May 27, 2025 at 3:04 PM
🧵6/10 🧩 Puzzle C builds planning capabilities. The model learns to generate a complete, multi-step retouching plan to enhance a photo, structuring its reasoning as a sequence of discrete issues and solutions for clarity and control.
🧵5/10 🧩 Puzzle B imparts aesthetic judgement. By ranking professionally edited photos against altered versions, the MLLM learns to recognize the visual characteristics of an optimally adjusted image for any given operation, building an internal aesthetic model.
May 27, 2025 at 3:04 PM
🧵5/10 🧩 Puzzle B imparts aesthetic judgement. By ranking professionally edited photos against altered versions, the MLLM learns to recognize the visual characteristics of an optimally adjusted image for any given operation, building an internal aesthetic model.
🧵4/10 🧩 Puzzle A builds an understanding of individual operations. The MLLM learns to map visual changes in before/after images to a specific tool and its precise parameter value, effectively learning the semantics of our procedural library.
May 27, 2025 at 3:04 PM
🧵4/10 🧩 Puzzle A builds an understanding of individual operations. The MLLM learns to map visual changes in before/after images to a specific tool and its precise parameter value, effectively learning the semantics of our procedural library.
🧵3/10 Our key recipe: MLLMs struggle to predict edit values directly. We solve this by generating rich textual reasoning for each puzzle ✍️. We then fine-tune MonetGPT on this data, creating a 'reasoning pathway' that enables it to regress final adjustment values.
May 27, 2025 at 3:04 PM
🧵3/10 Our key recipe: MLLMs struggle to predict edit values directly. We solve this by generating rich textual reasoning for each puzzle ✍️. We then fine-tune MonetGPT on this data, creating a 'reasoning pathway' that enables it to regress final adjustment values.
🧵2/10 MLLMs lack the visual understanding to plan edits. 🧠 So, we use expert photos as our ground truth and work backward, procedurally creating puzzles by assuming any change to an expert edit makes it less optimal
May 27, 2025 at 3:04 PM
🧵2/10 MLLMs lack the visual understanding to plan edits. 🧠 So, we use expert photos as our ground truth and work backward, procedurally creating puzzles by assuming any change to an expert edit makes it less optimal
Amazon came pretty late to India and we already had some homegrown companies like Flipkart which now competes with Amazon and is valued at $40B.
I think big tech's early access killed homegrown companies. China and to an extent South Korea (Naver) has some great tech companies because of barriers
I think big tech's early access killed homegrown companies. China and to an extent South Korea (Naver) has some great tech companies because of barriers
February 27, 2025 at 11:50 AM
Amazon came pretty late to India and we already had some homegrown companies like Flipkart which now competes with Amazon and is valued at $40B.
I think big tech's early access killed homegrown companies. China and to an extent South Korea (Naver) has some great tech companies because of barriers
I think big tech's early access killed homegrown companies. China and to an extent South Korea (Naver) has some great tech companies because of barriers
Who will tell the silicon valley tech bros that it wasn't them alone
February 7, 2025 at 8:34 PM
Who will tell the silicon valley tech bros that it wasn't them alone
Haha exactly what I did today!
January 24, 2025 at 5:27 PM
Haha exactly what I did today!
Have a great time in Seattle!
December 2, 2024 at 7:40 PM
Have a great time in Seattle!