Lab 2d: DASNet Inference¶

In this lab, you will learn how to:

• Load one Monterey Bay DAS tile from Hugging Face (AI4EPS/quakeflow_das), same flow as Lab 2a;
• Run DASNet for detection, classification, masks, and arrival-oriented picks;
• Inspect saved JSON and quicklook figures;
• Optionally run the same pipeline through predict.py from the shell.

References:

• Zhang, C., et al. (2026). "A deep learning framework for marine acoustic and seismic monitoring with distributed acoustic sensing." arXiv:2603.14844.
• Romanowicz, B., et al. (2023). "SeaFOAM: A year‐long DAS deployment in Monterey Bay, California." SRL, 94(5), 2348-2359.

---

Setup¶

In [10]:

import os
import subprocess

REPO_URL = "https://github.com/czh4ng/DASNet-workshop.git"
WORK_DIR = os.path.abspath(os.path.join(os.getcwd(), "DASNet-workshop"))

if not os.path.isdir(WORK_DIR):
    subprocess.run(["git", "clone", REPO_URL, WORK_DIR], check=True)
    print("Cloned to:", WORK_DIR)
else:
    print("Directory already exists:", WORK_DIR)

os.chdir(WORK_DIR)
print("cwd:", os.getcwd())

import os import subprocess REPO_URL = "https://github.com/czh4ng/DASNet-workshop.git" WORK_DIR = os.path.abspath(os.path.join(os.getcwd(), "DASNet-workshop")) if not os.path.isdir(WORK_DIR): subprocess.run(["git", "clone", REPO_URL, WORK_DIR], check=True) print("Cloned to:", WORK_DIR) else: print("Directory already exists:", WORK_DIR) os.chdir(WORK_DIR) print("cwd:", os.getcwd())

Cloned to: /content/DASNet-workshop/DASNet-workshop
cwd: /content/DASNet-workshop/DASNet-workshop

In [11]:

%pip install -q -r requirements.txt
%pip install -q huggingface_hub

%pip install -q -r requirements.txt %pip install -q huggingface_hub

In [12]:

import csv
import json
import subprocess
from pathlib import Path

import h5py
import matplotlib.pyplot as plt
import numpy as np
import torch
import urllib.request
from IPython.display import Image, display
from huggingface_hub import hf_hub_download

%matplotlib inline

DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print("Device:", DEVICE)

DS_PLOT = 2


def plot_das(ax, arr_ch_by_time, dt_s=None, **kwargs):
    arr = np.asarray(arr_ch_by_time)
    extent = None
    if dt_s is not None:
        nt = arr.shape[1]
        extent = [0, nt * float(dt_s), arr.shape[0], 0]
    ax.imshow(
        arr[::DS_PLOT, ::DS_PLOT],
        aspect="auto",
        interpolation="bilinear",
        extent=extent,
        **kwargs,
    )


def _to_numpy_pred(d):
    return {k: (v.detach().cpu().numpy() if torch.is_tensor(v) else v) for k, v in d.items()}

import csv import json import subprocess from pathlib import Path import h5py import matplotlib.pyplot as plt import numpy as np import torch import urllib.request from IPython.display import Image, display from huggingface_hub import hf_hub_download %matplotlib inline DEVICE = torch.device("cuda" if torch.cuda.is_available() else "cpu") print("Device:", DEVICE) DS_PLOT = 2 def plot_das(ax, arr_ch_by_time, dt_s=None, **kwargs): arr = np.asarray(arr_ch_by_time) extent = None if dt_s is not None: nt = arr.shape[1] extent = [0, nt * float(dt_s), arr.shape[0], 0] ax.imshow( arr[::DS_PLOT, ::DS_PLOT], aspect="auto", interpolation="bilinear", extent=extent, **kwargs, ) def _to_numpy_pred(d): return {k: (v.detach().cpu().numpy() if torch.is_tensor(v) else v) for k, v in d.items()}

Device: cuda

1. Load DAS Data¶

We use DAS data from SeaFOAM in Monterey Bay, California (Romanowicz et al., 2023). DASNet was trained on SeaFOAM. The deployment uses a ~52 km fiber with 5.2 m channel spacing at 200 Hz sampling.

Shallow parts of the cable are dominated by cultural and sea-surface noise, so we focus on the deep-water section (channel index > 7400).

Each workshop tile is an HDF5 file containing a 2D array of shape (nx, nt) — channels by time — plus metadata for the sample interval.

Set EVENT_ID in the next cell. The following figure shows raw strain rate, 2–10 Hz bandpass, and >10 Hz highpass.

In [13]:

HF_REPO = "AI4EPS/quakeflow_das"
EVENT_ID = "20231109T132510Z"

h5_path = hf_hub_download(
    HF_REPO,
    f"monterey_bay/data/{EVENT_ID}.h5",
    repo_type="dataset",
    local_dir="data/quakeflow_das",
)
SELECTED_H5 = str(Path(h5_path).resolve())
print("Cached path:", SELECTED_H5)

with h5py.File(SELECTED_H5, "r") as fp:
    data = fp["data"][:]
    attrs = dict(fp["data"].attrs)

dt_s = float(attrs.get("dt_s"))
begin_time = str(attrs.get("begin_time", ""))
nx, nt = data.shape
das_arr = data.astype(np.float32)

print(f"Event: {EVENT_ID}")
print(f"Shape: (nx={nx}, nt={nt})")
print(f"Sampling interval: {dt_s} s ({1/dt_s:.0f} Hz)")
print(f"Duration: {nt * dt_s:.1f} s")
if begin_time:
    print(f"Begin time: {begin_time}")

HF_REPO = "AI4EPS/quakeflow_das" EVENT_ID = "20231109T132510Z" h5_path = hf_hub_download( HF_REPO, f"monterey_bay/data/{EVENT_ID}.h5", repo_type="dataset", local_dir="data/quakeflow_das", ) SELECTED_H5 = str(Path(h5_path).resolve()) print("Cached path:", SELECTED_H5) with h5py.File(SELECTED_H5, "r") as fp: data = fp["data"][:] attrs = dict(fp["data"].attrs) dt_s = float(attrs.get("dt_s")) begin_time = str(attrs.get("begin_time", "")) nx, nt = data.shape das_arr = data.astype(np.float32) print(f"Event: {EVENT_ID}") print(f"Shape: (nx={nx}, nt={nt})") print(f"Sampling interval: {dt_s} s ({1/dt_s:.0f} Hz)") print(f"Duration: {nt * dt_s:.1f} s") if begin_time: print(f"Begin time: {begin_time}")

monterey_bay/data/20231109T132510Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

Cached path: /content/DASNet-workshop/DASNet-workshop/data/quakeflow_das/monterey_bay/data/20231109T132510Z.h5
Event: 20231109T132510Z
Shape: (nx=2845, nt=12000)
Sampling interval: 0.005000114440917969 s (200 Hz)
Duration: 60.0 s
Begin time: 2023-11-09T13:25:10.000000+00:00

In [14]:

from scipy.signal import sosfiltfilt

from dasnet.data import das as das_das

sos_bp = das_das._safe_design_sos_bandpass(dt_s, 2.0, 10.0, order=4)
sos_hp = das_das._safe_design_sos_highpass(dt_s, 10.0, order=4)
das_bp = sosfiltfilt(sos_bp, das_arr, axis=1).astype(np.float32)
das_hp = sosfiltfilt(sos_hp, das_arr, axis=1).astype(np.float32)

fig, axes = plt.subplots(3, 1, figsize=(10, 8))

titles = [
    "Raw strain rate",
    "2–10 Hz bandpass",
    ">10 Hz highpass",
]
arrays = [das_arr, das_bp, das_hp]

for ax, arr, title in zip(axes, arrays, titles):
    v = 2 * float(np.percentile(np.abs(arr), 95))
    plot_das(ax, arr, dt_s=dt_s, cmap="seismic", vmin=-v, vmax=v)
    ax.set_title(title)
    ax.set_xlabel("Time (s)")
    ax.set_ylabel("Channel index")

fig.suptitle(f"{EVENT_ID}", y=1.02)
plt.tight_layout()
plt.show()

from scipy.signal import sosfiltfilt from dasnet.data import das as das_das sos_bp = das_das._safe_design_sos_bandpass(dt_s, 2.0, 10.0, order=4) sos_hp = das_das._safe_design_sos_highpass(dt_s, 10.0, order=4) das_bp = sosfiltfilt(sos_bp, das_arr, axis=1).astype(np.float32) das_hp = sosfiltfilt(sos_hp, das_arr, axis=1).astype(np.float32) fig, axes = plt.subplots(3, 1, figsize=(10, 8)) titles = [ "Raw strain rate", "2–10 Hz bandpass", ">10 Hz highpass", ] arrays = [das_arr, das_bp, das_hp] for ax, arr, title in zip(axes, arrays, titles): v = 2 * float(np.percentile(np.abs(arr), 95)) plot_das(ax, arr, dt_s=dt_s, cmap="seismic", vmin=-v, vmax=v) ax.set_title(title) ax.set_xlabel("Time (s)") ax.set_ylabel("Channel index") fig.suptitle(f"{EVENT_ID}", y=1.02) plt.tight_layout() plt.show()

2. Run DASNet¶

DASNet outputs a class, confidence, box, and soft mask per detection; peaks along the mask summarize arrivals where applicable. The next cell downloads the workshop checkpoint, runs the selected tile (EVENT_ID above) through the model, and saves JSON plus a figure under pred_notebook/.

In [15]:

CKPT_URL = "https://github.com/czh4ng/DASNet-workshop/releases/download/v0.1.1/checkpoint.pth"
CKPT_PATH = Path(WORK_DIR) / "checkpoint.pth"
if not CKPT_PATH.exists():
    urllib.request.urlretrieve(CKPT_URL, CKPT_PATH)

from lab2d_inference_helpers import (
    build_dasnet_model,
    default_device,
    extract_peaks_for_instances,
    filter_by_score,
    forward_raw,
    load_checkpoint,
    make_infer_dataloader,
    plot_das_predictions,
    postprocess_batch,
    save_predictions_json,
)

RESIZE_SCALE = 0.5
MIN_PROB = 0.8

device = default_device()
model = build_dasnet_model()
load_checkpoint(model, str(CKPT_PATH), device)

loader, _ = make_infer_dataloader(
    [SELECTED_H5],
    batch_size=1,
    num_workers=0,
    resize_scale=RESIZE_SCALE,
    storage_backend="local",
)
images, names = next(iter(loader))

raw_outputs = forward_raw(model, list(images), device)
_, processed = postprocess_batch(names, raw_outputs)
p0 = processed[0]
selected = filter_by_score(p0, MIN_PROB)
peak_points_list, peak_scores_list = extract_peaks_for_instances(selected)

RESULT_DIR = Path(WORK_DIR) / "pred_notebook"
FIG_DIR = RESULT_DIR / "figures_dasnet"
RESULT_DIR.mkdir(parents=True, exist_ok=True)
FIG_DIR.mkdir(parents=True, exist_ok=True)

fn0 = names[0]
save_predictions_json(
    fn0,
    _to_numpy_pred(selected),
    peak_points_list,
    peak_scores_list,
    str(RESULT_DIR),
    resize_scale=RESIZE_SCALE,
)
fig_path = FIG_DIR / (Path(fn0).stem + ".jpg")
plot_das_predictions(
    images[0].cpu().numpy(),
    _to_numpy_pred(selected),
    str(fig_path),
    score_threshold=MIN_PROB,
)

print("Instances kept:", len(selected["scores"]), "| JSON + figure →", RESULT_DIR)

CKPT_URL = "https://github.com/czh4ng/DASNet-workshop/releases/download/v0.1.1/checkpoint.pth" CKPT_PATH = Path(WORK_DIR) / "checkpoint.pth" if not CKPT_PATH.exists(): urllib.request.urlretrieve(CKPT_URL, CKPT_PATH) from lab2d_inference_helpers import ( build_dasnet_model, default_device, extract_peaks_for_instances, filter_by_score, forward_raw, load_checkpoint, make_infer_dataloader, plot_das_predictions, postprocess_batch, save_predictions_json, ) RESIZE_SCALE = 0.5 MIN_PROB = 0.8 device = default_device() model = build_dasnet_model() load_checkpoint(model, str(CKPT_PATH), device) loader, _ = make_infer_dataloader( [SELECTED_H5], batch_size=1, num_workers=0, resize_scale=RESIZE_SCALE, storage_backend="local", ) images, names = next(iter(loader)) raw_outputs = forward_raw(model, list(images), device) _, processed = postprocess_batch(names, raw_outputs) p0 = processed[0] selected = filter_by_score(p0, MIN_PROB) peak_points_list, peak_scores_list = extract_peaks_for_instances(selected) RESULT_DIR = Path(WORK_DIR) / "pred_notebook" FIG_DIR = RESULT_DIR / "figures_dasnet" RESULT_DIR.mkdir(parents=True, exist_ok=True) FIG_DIR.mkdir(parents=True, exist_ok=True) fn0 = names[0] save_predictions_json( fn0, _to_numpy_pred(selected), peak_points_list, peak_scores_list, str(RESULT_DIR), resize_scale=RESIZE_SCALE, ) fig_path = FIG_DIR / (Path(fn0).stem + ".jpg") plot_das_predictions( images[0].cpu().numpy(), _to_numpy_pred(selected), str(fig_path), score_threshold=MIN_PROB, ) print("Instances kept:", len(selected["scores"]), "| JSON + figure →", RESULT_DIR)

Instances kept: 6 | JSON + figure → /content/DASNet-workshop/DASNet-workshop/pred_notebook

3. Check output¶

Open the quicklook figure and JSON for the same tile as EVENT_ID / Run DASNet. Each instance stores the box and (when present) pick coordinates in full-resolution channel/time units after undoing the resize, together with class id, score, and per-pick mask scores.

In [16]:

if fig_path.exists():
    display(Image(filename=str(fig_path)))
else:
    print("No figure at", fig_path)

json_path = RESULT_DIR / (Path(fn0).stem + ".json")
with open(json_path) as f:
    payload = json.load(f)

view = json.loads(json.dumps(payload))
for inst in view.get("instances", []):
    for key in ("picks", "pick_scores"):
        if key in inst and len(inst[key]) > 5:
            n = len(inst[key])
            inst[key] = inst[key][:5] + [f"... ({n} total)"]

print(json.dumps(view, indent=2))

if fig_path.exists(): display(Image(filename=str(fig_path))) else: print("No figure at", fig_path) json_path = RESULT_DIR / (Path(fn0).stem + ".json") with open(json_path) as f: payload = json.load(f) view = json.loads(json.dumps(payload)) for inst in view.get("instances", []): for key in ("picks", "pick_scores"): if key in inst and len(inst[key]) > 5: n = len(inst[key]) inst[key] = inst[key][:5] + [f"... ({n} total)"] print(json.dumps(view, indent=2))

{
  "file_name": "20231109T132510Z.h5",
  "resize_scale": 0.5,
  "instances": [
    {
      "box": [
        164.8438720703125,
        914.1192626953125,
        2843.15625,
        3053.880859375
      ],
      "score": 0.9995997548103333,
      "label": 3,
      "label_name": "Fin whale",
      "picks": [
        [
          168.0,
          2770.0
        ],
        [
          170.0,
          2768.0
        ],
        [
          172.0,
          2766.0
        ],
        [
          174.0,
          2762.0
        ],
        [
          176.0,
          2760.0
        ],
        "... (1338 total)"
      ],
      "pick_scores": [
        0.529354989528656,
        0.5951864719390869,
        0.6569203734397888,
        0.7130587100982666,
        0.7625828981399536,
        "... (1338 total)"
      ]
    },
    {
      "box": [
        167.2171630859375,
        9504.1396484375,
        2840.78271484375,
        11615.8603515625
      ],
      "score": 0.9994791150093079,
      "label": 3,
      "label_name": "Fin whale",
      "picks": [
        [
          170.0,
          11370.0
        ],
        [
          172.0,
          11368.0
        ],
        [
          174.0,
          11366.0
        ],
        [
          176.0,
          11364.0
        ],
        [
          178.0,
          11362.0
        ],
        "... (1335 total)"
      ],
      "pick_scores": [
        0.5304351449012756,
        0.5961430668830872,
        0.6576252579689026,
        0.7133526802062988,
        0.762298047542572,
        "... (1335 total)"
      ]
    },
    {
      "box": [
        301.1693115234375,
        5273.41552734375,
        2834.83056640625,
        7270.58447265625
      ],
      "score": 0.9994751811027527,
      "label": 3,
      "label_name": "Fin whale",
      "picks": [
        [
          304.0,
          6996.0
        ],
        [
          306.0,
          6992.0
        ],
        [
          308.0,
          6990.0
        ],
        [
          310.0,
          6986.0
        ],
        [
          312.0,
          6984.0
        ],
        "... (1263 total)"
      ],
      "pick_scores": [
        0.5291606783866882,
        0.5946370363235474,
        0.6558182835578918,
        0.7112131118774414,
        0.7597672343254089,
        "... (1263 total)"
      ]
    },
    {
      "box": [
        501.03662109375,
        8800.083984375,
        2844.0,
        10271.916015625
      ],
      "score": 0.9977980852127075,
      "label": 3,
      "label_name": "Fin whale",
      "picks": [
        [
          504.0,
          10106.0
        ],
        [
          506.0,
          10106.0
        ],
        [
          508.0,
          10104.0
        ],
        [
          510.0,
          10104.0
        ],
        [
          512.0,
          10102.0
        ],
        "... (1170 total)"
      ],
      "pick_scores": [
        0.5391365885734558,
        0.603882908821106,
        0.6633433699607849,
        0.7158526182174683,
        0.7603036761283875,
        "... (1170 total)"
      ]
    },
    {
      "box": [
        557.8836669921875,
        207.14825439453125,
        2844.0,
        1568.851806640625
      ],
      "score": 0.9948232173919678,
      "label": 3,
      "label_name": "Fin whale",
      "picks": [
        [
          562.0,
          1370.0
        ],
        [
          564.0,
          1370.0
        ],
        [
          566.0,
          1370.0
        ],
        [
          568.0,
          1370.0
        ],
        [
          570.0,
          1370.0
        ],
        "... (1141 total)"
      ],
      "pick_scores": [
        0.5275543928146362,
        0.5851976275444031,
        0.6383750438690186,
        0.6860569715499878,
        0.727648913860321,
        "... (1141 total)"
      ]
    },
    {
      "box": [
        713.1239013671875,
        4532.55126953125,
        2844.0,
        5963.44873046875
      ],
      "score": 0.9921293258666992,
      "label": 3,
      "label_name": "Fin whale",
      "picks": [
        [
          716.0,
          5728.0
        ],
        [
          718.0,
          5728.0
        ],
        [
          720.0,
          5666.0
        ],
        [
          722.0,
          5666.0
        ],
        [
          724.0,
          5666.0
        ],
        "... (1064 total)"
      ],
      "pick_scores": [
        0.5163047909736633,
        0.5751291513442993,
        0.629677951335907,
        0.6799686551094055,
        0.7230713367462158,
        "... (1064 total)"
      ]
    }
  ]
}

Optional: predict.py from the command line¶

The same end-to-end inference is available as a script. Edit CLI_EVENT_IDS (one or more compact UTC IDs), download each file with hf_hub_download, write a data_list.csv with a path column, and run predict.py once. Outputs go to pred_cli/ (JSON + figures_dasnet/).

The following cell shows quicklook figures in data_list.csv order. Set FIG_PREVIEW_N to None for all rows, or a positive integer for only the first n.

In [17]:

from tqdm.auto import tqdm

HF_REPO_BATCH = "AI4EPS/quakeflow_das"
CLI_LOCAL_DIR = Path(WORK_DIR) / "data" / "quakeflow_das"

CLI_EVENT_IDS = [
    "20231116T130210Z",
    "20231110T105410Z",
    "20231124T104510Z",
    "20231130T194610Z",
    "20240204T004410Z",
    "20240428T100822Z",
    "20220828T062358Z",
    "20220828T140758Z",
    "20230407T074209Z",
]

pred_cli = Path(WORK_DIR) / "pred_cli"
pred_cli.mkdir(parents=True, exist_ok=True)
csv_path = pred_cli / "data_list.csv"

paths_for_cli = []
for eid in tqdm(CLI_EVENT_IDS, desc="HF fetch (batch)", unit="file"):
    p = hf_hub_download(
        HF_REPO_BATCH,
        f"monterey_bay/data/{eid}.h5",
        repo_type="dataset",
        local_dir=str(CLI_LOCAL_DIR),
    )
    paths_for_cli.append(str(Path(p).resolve()))

with open(csv_path, "w", newline="") as fp:
    w = csv.writer(fp)
    w.writerow(["path"])
    for p in paths_for_cli:
        w.writerow([p])

print(f"data_list.csv: {len(paths_for_cli)} rows -> {csv_path.resolve()}")

predict_py = Path(WORK_DIR) / "predict.py"
base = [
    str(predict_py),
    "--model",
    "dasnet",
    "--resume",
    str(CKPT_PATH),
    "--data_list",
    str(csv_path),
    "--batch_size",
    "1",
    "--min_prob",
    str(MIN_PROB),
    "--result_path",
    str(pred_cli),
    "--plot_figure",
    "--workers",
    "0",
]
if DEVICE.type == "cpu":
    cmd = ["python"] + base + ["--device", "cpu"]
else:
    cmd = ["python"] + base

print("Running:", " ".join(cmd))
proc = subprocess.run(cmd, cwd=str(WORK_DIR), capture_output=True, text=True)
print(proc.stdout)
if proc.stderr:
    print(proc.stderr)
proc.check_returncode()

n_json = len(list(pred_cli.glob("*.json")))
print(f"Finished. JSON count under pred_cli: {n_json}")

from tqdm.auto import tqdm HF_REPO_BATCH = "AI4EPS/quakeflow_das" CLI_LOCAL_DIR = Path(WORK_DIR) / "data" / "quakeflow_das" CLI_EVENT_IDS = [ "20231116T130210Z", "20231110T105410Z", "20231124T104510Z", "20231130T194610Z", "20240204T004410Z", "20240428T100822Z", "20220828T062358Z", "20220828T140758Z", "20230407T074209Z", ] pred_cli = Path(WORK_DIR) / "pred_cli" pred_cli.mkdir(parents=True, exist_ok=True) csv_path = pred_cli / "data_list.csv" paths_for_cli = [] for eid in tqdm(CLI_EVENT_IDS, desc="HF fetch (batch)", unit="file"): p = hf_hub_download( HF_REPO_BATCH, f"monterey_bay/data/{eid}.h5", repo_type="dataset", local_dir=str(CLI_LOCAL_DIR), ) paths_for_cli.append(str(Path(p).resolve())) with open(csv_path, "w", newline="") as fp: w = csv.writer(fp) w.writerow(["path"]) for p in paths_for_cli: w.writerow([p]) print(f"data_list.csv: {len(paths_for_cli)} rows -> {csv_path.resolve()}") predict_py = Path(WORK_DIR) / "predict.py" base = [ str(predict_py), "--model", "dasnet", "--resume", str(CKPT_PATH), "--data_list", str(csv_path), "--batch_size", "1", "--min_prob", str(MIN_PROB), "--result_path", str(pred_cli), "--plot_figure", "--workers", "0", ] if DEVICE.type == "cpu": cmd = ["python"] + base + ["--device", "cpu"] else: cmd = ["python"] + base print("Running:", " ".join(cmd)) proc = subprocess.run(cmd, cwd=str(WORK_DIR), capture_output=True, text=True) print(proc.stdout) if proc.stderr: print(proc.stderr) proc.check_returncode() n_json = len(list(pred_cli.glob("*.json"))) print(f"Finished. JSON count under pred_cli: {n_json}")

HF fetch (batch):   0%|          | 0/9 [00:00<?, ?file/s]

monterey_bay/data/20231116T130210Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20231110T105410Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20231124T104510Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20231130T194610Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20240204T004410Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20240428T100822Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20220828T062358Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20220828T140758Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

monterey_bay/data/20230407T074209Z.h5:   0%|          | 0.00/137M [00:00<?, ?B/s]

data_list.csv: 9 rows -> /content/DASNet-workshop/DASNet-workshop/pred_cli/data_list.csv
Running: python /content/DASNet-workshop/DASNet-workshop/predict.py --model dasnet --resume /content/DASNet-workshop/DASNet-workshop/checkpoint.pth --data_list /content/DASNet-workshop/DASNet-workshop/pred_cli/data_list.csv --batch_size 1 --min_prob 0.8 --result_path /content/DASNet-workshop/DASNet-workshop/pred_cli --plot_figure --workers 0
Not using distributed mode
Namespace(model='dasnet', resume='/content/DASNet-workshop/DASNet-workshop/checkpoint.pth', device='cuda', batch_size=1, workers=0, amp=False, distributed=False, world_size=1, dist_url='env://', data_list='/content/DASNet-workshop/DASNet-workshop/pred_cli/data_list.csv', object='default', key_path=None, result_path='/content/DASNet-workshop/DASNet-workshop/pred_cli', min_prob=0.8, plot_figure=True, skip_existing=False, use_deterministic_algorithms=False)
Total file number from list: 9
Loading checkpoint: /content/DASNet-workshop/DASNet-workshop/checkpoint.pth


Predicting:   0%|          | 0/9 [00:00<?, ?it/s]
Predicting:  11%|█         | 1/9 [02:34<20:32, 154.04s/it]
Predicting:  22%|██▏       | 2/9 [03:41<12:01, 103.10s/it]
Predicting:  33%|███▎      | 3/9 [04:04<06:38, 66.36s/it] 
Predicting:  44%|████▍     | 4/9 [06:38<08:25, 101.14s/it]
Predicting:  56%|█████▌    | 5/9 [07:01<04:51, 72.91s/it] 
Predicting:  67%|██████▋   | 6/9 [07:24<02:47, 55.87s/it]
Predicting:  78%|███████▊  | 7/9 [07:48<01:30, 45.49s/it]
Predicting:  89%|████████▉ | 8/9 [08:40<00:47, 47.71s/it]
Predicting: 100%|██████████| 9/9 [09:03<00:00, 39.84s/it]
Predicting: 100%|██████████| 9/9 [09:03<00:00, 60.37s/it]

Finished. JSON count under pred_cli: 9

In [18]:

FIG_PREVIEW_N = None  # None = show all CSV rows; set int n to show only the first n

pred_cli = Path(WORK_DIR) / "pred_cli"
cli_fig_dir = pred_cli / "figures_dasnet"

with open(pred_cli / "data_list.csv", newline="") as fp:
    rows_all = list(csv.DictReader(fp))
if not rows_all:
    raise FileNotFoundError("Run the previous cell first (data_list.csv missing or empty).")

if FIG_PREVIEW_N is None:
    rows = rows_all
else:
    rows = rows_all[: max(0, int(FIG_PREVIEW_N))]

n_all = len(rows_all)
n = len(rows)
print(f"Showing {n} of {n_all} data_list.csv row(s) (FIG_PREVIEW_N={FIG_PREVIEW_N!r}).")

for i, row in enumerate(rows):
    h5_name = Path(row["path"]).name
    fig_path = cli_fig_dir / (h5_name + ".jpg")
    print(f"[{i + 1}/{n}] {h5_name}")
    if fig_path.exists():
        display(Image(filename=str(fig_path)))
        print(fig_path.resolve())
    else:
        print("  (figure missing)", fig_path.resolve())
    print()

FIG_PREVIEW_N = None # None = show all CSV rows; set int n to show only the first n pred_cli = Path(WORK_DIR) / "pred_cli" cli_fig_dir = pred_cli / "figures_dasnet" with open(pred_cli / "data_list.csv", newline="") as fp: rows_all = list(csv.DictReader(fp)) if not rows_all: raise FileNotFoundError("Run the previous cell first (data_list.csv missing or empty).") if FIG_PREVIEW_N is None: rows = rows_all else: rows = rows_all[: max(0, int(FIG_PREVIEW_N))] n_all = len(rows_all) n = len(rows) print(f"Showing {n} of {n_all} data_list.csv row(s) (FIG_PREVIEW_N={FIG_PREVIEW_N!r}).") for i, row in enumerate(rows): h5_name = Path(row["path"]).name fig_path = cli_fig_dir / (h5_name + ".jpg") print(f"[{i + 1}/{n}] {h5_name}") if fig_path.exists(): display(Image(filename=str(fig_path))) print(fig_path.resolve()) else: print(" (figure missing)", fig_path.resolve()) print()

Showing 9 of 9 data_list.csv row(s) (FIG_PREVIEW_N=None).
[1/9] 20231116T130210Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20231116T130210Z.h5.jpg

[2/9] 20231110T105410Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20231110T105410Z.h5.jpg

[3/9] 20231124T104510Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20231124T104510Z.h5.jpg

[4/9] 20231130T194610Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20231130T194610Z.h5.jpg

[5/9] 20240204T004410Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20240204T004410Z.h5.jpg

[6/9] 20240428T100822Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20240428T100822Z.h5.jpg

[7/9] 20220828T062358Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20220828T062358Z.h5.jpg

[8/9] 20220828T140758Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20220828T140758Z.h5.jpg

[9/9] 20230407T074209Z.h5

/content/DASNet-workshop/DASNet-workshop/pred_cli/figures_dasnet/20230407T074209Z.h5.jpg