🏠 HouseObj_VQA

Household Object Visual Question Answering Dataset

📌 Dataset Details

🧾 Dataset Description

HouseObj_VQA is a curated Visual Question Answering (VQA) dataset designed to evaluate the capability of Vision-Language Models (VLMs) in identifying and reasoning about household objects.

The dataset focuses exclusively on commonly used household items, such as cream containers, coconut oil bottles, carrom boards, utensils, and similar objects. Each image is paired with one or more natural-language questions and corresponding answers that test object recognition, material identification, and brand/manufacturer awareness.

• 👤 Curated by: Department of Electronics and Communication Engineering, Bapatla Engineering College, Bapatla
• 🏫 Institutional Contribution: Academic research and evaluation
• 🌐 Shared by: Chandrabhuma (Hugging Face)
• 🗣️ Language: English
• 📜 License: Apache-2.0

---

🔗 Dataset Sources

• 🧠 Repository: https://huggingface.co/datasets/chandrabhuma/HouseObj_VQA
• 📄 Paper: Not available (dataset contribution)
• 🧪 Demo: Intended for VLM benchmarking and evaluation

---

🎯 Uses

✅ Direct Use

This dataset is intended for:

• 🧠 Evaluating Vision-Language Models (VLMs)
• 🔍 Testing household object recognition
• 🧱 Assessing material identification
• 🏷️ Brand / manufacturer awareness in visual reasoning
• 📊 Benchmarking open-ended VQA performance

Typical example questions:

• ❓ What is the object in the image? → Carrom board
• ❓ What is the material of the object? → Steel
• ❓ Identify the manufacturer of the object. → Cipla

---

🚫 Out-of-Scope Use

This dataset is not suitable for:

• ❌ Surveillance or personal identification
• ❌ Medical or safety-critical decision making
• ❌ Fine-grained industrial inspection
• ❌ Non-household object classification

---

🧱 Dataset Structure

Each data sample contains:

Field	Description
🆔 id	Unique sample identifier
🖼️ image_name	Image file name
❓ question	Natural-language question
✅ answer	Ground-truth answer
🖼️ image	Household object image

📊 Dataset Split

• 🟢 Training set: 800 images
• 🔵 Test set: 200 images
• 📦 Total images: 1000

---

🛠️ Dataset Creation

🎯 Curation Rationale

The dataset was created to address the lack of household-object-centric VQA benchmarks, especially for evaluating how well modern VLMs understand everyday objects encountered in real-world domestic environments.

---

📥 Source Data

🔄 Data Collection and Processing

• 📸 Images captured manually using consumer-grade cameras
• 🧹 Quality filtering to remove blurred or ambiguous images
• ✍️ Manual question-answer creation
• 🔍 Verification for correctness and consistency

---

👥 Who are the source data producers?

The data was collected and annotated by undergraduate students under academic supervision as part of a research initiative.

---

🏷️ Annotations

✏️ Annotation Process

• 👁️ Manual visual inspection of each image
• 🧠 Question design targeting object identity, material, and brand
• 📋 Answer verification by multiple contributors

---

👤 Annotators

👩‍💻 Dr.B.Chandra Mohan (Prof, ECE Dept, Bapatla Engineering College, Bapatla-522102) 📧 chandrabhuma@gmail.com

👩‍💻 Sri K.Sriharsha (Asst.Prof, ECE Dept, Bapatla Engineering College, Bapatla-522102) 📧 sriharsha4b2@gmail.com

• 👩‍💻 Naralasetty Mounika (UG Student , ECE Dept, Bapatla Engineering College, Bapatla-522102) 📧 naralasettymounika@gmail.com

• 👩‍💻 Koneti Surya Mounika (UG Student, ECE Dept, Bapatla Engineering College, Bapatla-522102) 📧 mounikasurya860@gmail.com

• 👨‍💻 Medisetti Harsha Vardhan (UG Student, ECE Dept, Bapatla Engineering College, Bapatla-522102) 📧 medisettiharshavardhan19@gmail.com

• 👨‍💻 Kurmala Siva Kumar (UG Student, ECE Dept, Bapatla Engineering College, Bapatla-522102) 📧 kumarsiva3951@gmail.com

---

🔐 Personal and Sensitive Information

🚫 The dataset does not contain:

• Personal identities
• Faces
• Private locations
• Sensitive or confidential information

All images depict non-personal household objects only.

---

⚠️ Bias, Risks, and Limitations

• ⚖️ Limited to household objects available in local environments
• 🌍 Cultural and regional object bias may exist
• 🧠 Brand recognition depends on visual clarity

---

📝 Recommendations

Users should:

• Combine this dataset with other object-centric datasets for generalization
• Avoid over-interpreting brand-recognition results
• Treat this dataset as an evaluation benchmark, not exhaustive coverage

---

📚 Citation

📌 BibTeX

@dataset{houseobj_vqa_2025,
  title={HouseObj\_VQA: A Household Object Visual Question Answering Dataset},
  author={Chandra mohan bhuma, sri harsha k, Naralasetty, Mounika and Koneti, Surya Mounika and Medisetti, Harsha Vardhan and Kurmala, Siva Kumar},
  year={2025},
  publisher={Hugging Face}
}