Imageomics / Imageomics/FuncaPalooza-2025
Trait Detection & Species Identification: A Tropical Seedling Case Study
- Dominant language
- No language data
- Stars
- 5
- Forks
- 0
- PR merge metrics
- No merged PRs in 30d
Description
## Trait Detection & Species Identification: A Tropical Seedling Case Study
I’d like to explore how images of tropical forest seedlings (young trees) can be used for species identification, extracting diagnostic traits (e.g., serrated leaf margins) and architectural features (e.g., leaf arrangement, branching) with image-based tools. While my focus is on seedlings, the approach could be relevant to other groups where identifying early life stages or detecting key traits is challenging.
### Dataset overview
- ~10,000 RGB ground-based field images of established tropical tree seedlings (young trees under ~1m tall)
- 195 species from multiple sites across Panama
- 5–6 angles per individual, collected under natural light (capturing leaf damage, lighting variation, and plant architecture)
- Expert-verified species IDs linked to metadata (site, observer, unique code)
These are live plants photographed in natural field conditions, capturing realistic variation in leaf damage, imperfect lighting, and natural growth patterns, unlike the clean, controlled images often used in herbariums. The diversity of tropical forests is high, and even expert botanists often find in situ seedling identification challenging. This makes the dataset a particularly strong test case for exploring approaches beyond standard model-based classification.
I’m open to any directions that work with this seedling image set, but I’m particularly curious about approaches that move further from the usual model-based classification and lean towards lines close to imitating how a botanist works. This is why I find the idea of focusing on diagnostic traits especially appealing.
Seedlings play a critical role in forest regeneration, making accurate identification essential for ecological monitoring and conservation.
Looking forward to feedback and more discussions!
Contributor guide
No contributing guide indexed for this repository
Research direction
No repository files, tests, or entry points are named. Start by reviewing the dataset description and narrowing the exploration to a specific trait-detection or species-identification goal; define the approach, evaluation criteria, and completion scope before implementation.
Written by the indexing model from the issue text.
Assessment
- Domain
- computer-vision, machine-learning
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Stale
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100