ecsam_system_main.py

# ECSAM Main Orchestrator
from edge_device import EdgeDevice
from cloud_server import CloudServer
from ecsam_modules import (
    ECSAM_Context_Mgmt, ECSAM_VP_Trans, ECSAM_ImgEnc, ECSAM_Dec, ECSAM_Cloud_Reg, ECSAM_Cloud_Offloading
)
import numpy as np

if __name__ == "__main__":
    # Initialize edge and cloud
    edge = EdgeDevice()
    cloud = CloudServer()

    # 1. Context management (prompt encoding/session)
    context_mgmt = ECSAM_Context_Mgmt()
    prompt = {'type': 'point', 'coords': (100, 100)}
    context_mgmt.encode_prompt(prompt)
    print("[ECSAM] Context management and prompt encoding done.")

    # 2. Visual prompt transformation
    vp_trans = ECSAM_VP_Trans()
    # Simulate a list of prompt objects (with dummy embeddings)
    from visual_prompt_transformer import Prompt
    prompts = [Prompt(embedding=np.random.rand(16), prompt_type='point', metadata=(i*10, i*10)) for i in range(3)]
    transformed_prompts = vp_trans.transform(prompts, latency_constraint=2)
    print(f"[ECSAM] Visual prompt transformation done. {len(transformed_prompts)} prompts after transformation.")

    # 3. Image encoding with partitioning
    img_enc = ECSAM_ImgEnc()
    dummy_image = np.random.randint(0, 255, (224, 224, 3), dtype=np.uint8)
    split_point = img_enc.model_partition(layers=list(range(100)), edge_costs=[1.0]*100, latency_constraint=60)
    img_emb = img_enc.encode_image(dummy_image, split_point=split_point)
    print(f"[ECSAM] Image encoding done. Split point: {split_point}")

    # 4. Decoding
    dec = ECSAM_Dec()
    # Use dummy embeddings for demonstration
    prompt_emb = np.random.rand(1, 16)
    image_emb = np.random.rand(1, 32)
    result = dec.decode(prompt_emb, image_emb)
    print(f"[ECSAM] Decoding done. Result: {result}")

    # 5. Cloud registration
    cloud_reg = ECSAM_Cloud_Reg()
    cloud_reg.register_edge(edge_id="edge-001")
    print("[ECSAM] Cloud registration done.")

    # 6. Cloud offloading
    cloud_offload = ECSAM_Cloud_Offloading()
    cloud_offload.offload_inference(data={'image': dummy_image})
    print("[ECSAM] Cloud offloading done.")